WO2023011294A1

WO2023011294A1 - Signal processing method and apparatus, and processing chip and signal transmission system

Info

Publication number: WO2023011294A1
Application number: PCT/CN2022/108339
Authority: WO
Inventors: 尚冬冬; 李星; 周蒙
Original assignee: 华为技术有限公司
Priority date: 2021-08-04
Filing date: 2022-07-27
Publication date: 2023-02-09
Also published as: CN115706688A

Abstract

A signal processing method and apparatus, and a processing chip and a signal transmission system. The signal processing apparatus comprises an equalization module, a rough decision module, a filtering module, a maximum likelihood sequence estimation (MLSE) module and a mapping module, wherein the equalization module is used for performing equalization processing on an initial signal to obtain an equalization signal; the rough decision module is used for determining dimension reduction position information and a dimension reduction signal according to the equalization signal, the number of signal states of the dimension reduction signal being less than the number of signal states of the equalization signal, and the dimension reduction signal being unrelated to a modulation format; the filtering module is used for performing filtering processing on the dimension reduction signal to obtain a filtering signal; the MLSE module is used for determining a target path branch of the filtering signal; and the mapping module is used for performing, according to the dimension reduction position information, mapping on the target path branch to obtain a target signal. By means of the present application, the resource consumption of an MLSE module is reduced.

Description

Signal processing method and device, processing chip, signal transmission system

This application claims the priority of a Chinese patent application with an application date of August 4, 2021, an application number of 202110890844.8, and an application title of "Signal Processing Method and Device, Processing Chip, and Signal Transmission System", the entire contents of which are incorporated herein by reference. In this application.

technical field

The present application relates to the technical field of communications, and in particular to a signal processing method and device, a processing chip, and a signal transmission system.

Background technique

A signal transmission system generally includes a signal sending device and a signal receiving device, the signal sending device and the signal receiving device are connected through a transmission link (link), and the signal sending device sends a signal to the signal receiving device through the transmission link. During the signal transmission process, limited by the bandwidth of each device in the signal transmission system, the signal usually has intersymbol intrference (ISI). In order to ensure that the signal receiving device recovers correct data from the received signal, after receiving the signal, the signal receiving device usually uses an equalization algorithm to equalize the signal to eliminate or weaken the ISI of the signal. The MLSE algorithm is an equalization algorithm with better performance, so the MLSE algorithm is usually used to equalize the signal.

The signal receiving device usually includes a digital signal processing (digital signal processing, DSP) chip, and the DSP chip performs equalization processing on the signal. Exemplarily, the DSP chip includes a feed forward equalization (feed forward equalization, FFE), a post filter (post filter) and an MLSE module connected in sequence, and a path branch calculation module respectively connected to the MLSE module and the post filter. The FFE equalizes the received signal to obtain an equalized signal. The equalized signal has no ISI and is nonlinear, but the equalized signal contains colored noise. The post-filter filters the equalized signal to obtain a filtered signal. During the process of filtering the equalized signal, the post-filter converts the colored noise in the equalized signal into white noise, and introduces a variable noise into the equalized signal. controlled ISI. The path branch calculation module calculates the number of path branches of the filtered signal (that is, the number of possible sequences of the filtered signal) according to the filter coefficients of the post filter. The MLSE module performs MLSE processing on the filtered signal according to the number of path branches of the filtered signal to obtain an optimal solution. Among them, the complexity of the MLSE algorithm depends on the number of path branches input to the MLSE module. Assuming that the number of states (that is, the number of levels) of the filtered signal is M, and the modulation order of the post-filter is L, the number of path branches input to the MLSE module is M ^L . That is, the number of path branches input to the MLSE module is positively correlated with the modulation order of the post-filter and the number of states (ie, the number of levels) of the filtered signal. The higher the modulation order of the post-filter, the higher the number of states of the filtered signal. The more, the higher the complexity of the MLSE algorithm.

In order to reduce the complexity of the MLSE algorithm and reduce the resource consumption of the MLSE process, current DSP chips usually use a second-order post-filter (that is, the modulation order of the post-filter is 2) for signal filtering. However, when a 2-order post-filter is used, the resource consumption of the MLSE process is still relatively large. For example, for a 4-level pulse amplitude modulation (phase amplitude modulation, PAM) signal (abbreviated as a PAM4 signal, the number of states of the PAM4 signal is 4), when a 2-order post-filter is used in the DSP chip, the path branch of the input MLSE module The number is 16 ( ^ML = 4 ² = 16), and the resource occupied by the MLSE module to perform MLSE processing on the filtered signal output by the post filter according to the path branch number 16 is as high as 1 million logic gate circuits, so the MLSE processing process Resource consumption is large.

Contents of the invention

The present application provides a signal processing method and device, a processing chip, and a signal transmission system, which help reduce resource consumption of MLSE. The technical scheme of the application is as follows:

In a first aspect, a signal processing device is provided, and the signal processing device includes: an equalization module, a rough decision module, a filter module, an MLSE module, and a mapping module connected in sequence, and the rough decision module is also connected to the mapping module. The equalization module is used to equalize the initial signal input to the equalization module to obtain an equalized signal. The rough judgment module is used to determine the dimensionality reduction position information and the dimensionality reduction signal according to the balanced signal, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the balanced signal, and the dimensionality reduction signal includes the dimensionality reduction at each moment in n moments The sub-signal, the dimension reduction signal has nothing to do with the modulation format, the dimension reduction position information includes the position identification value corresponding to each time in the n times, and the position identification value corresponding to each time is used to indicate the dimension reduction sub-signal at each time In the corresponding position in the equalized signal, n is a positive integer. The filtering module is used for filtering the dimensionality reduction signal to obtain the filtered signal. The MLSE module is used to determine the target path branch (eg, the optimal path branch) of the filtered signal. The mapping module is used to map the target path branch of the filtered signal to obtain the target signal (such as the optimal signal) according to the dimensionality reduction position information. The number of signal states of the target signal is equal to the number of signal states of the balanced signal. Each of the target signals The signal states are the same as or correspond to the respective signal states of the equalized signal.

Among them, the equalized signal has no ISI and is nonlinear, but the equalized signal contains colored noise, so the dimensionality reduction signal contains colored noise. In the process of filtering the dimensionality reduction signal, the filtering module converts the colored noise in the dimensionality reduction signal into white noise, and introduces controllable ISI into the dimensionality reduction signal, so the filtered signal contains white noise and controllable ISI.

Wherein, the target path branch output by the MLSE module includes decision values at each of the n time moments, and the decision values in the target path branch may be hard values or soft values. If the decision value in this target path branch is a hard value, then the individual signal states of the target signal are identical to the individual signal states of the equalized signal. If the decision value in the target path branch is a soft value, the respective signal states of the target signal correspond to the respective signal states of the equalized signal.

In the technical solution provided by this application, since the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalization signal, so the filtered signal input to the MLSE module has nothing to do with the modulation format, and the filtered signal The number of signal states is less than the number of signal states of the dimensionality reduction signal, and the MLSE module needs to calculate fewer path branches in the process of determining the target path branch of the filtered signal, and the process of determining the target path branch of the filtered signal by the MLSE module is the same as The modulation format is irrelevant, which helps to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module.

Optionally, for the corresponding position of the dimensionality reduction sub-signal at each moment in the equalized signal, the equalized signal has two state positions adjacent to the corresponding position, the corresponding position is located between the two state positions, the The Euclidean distance between the corresponding position and the two state positions is equal, each state position corresponds to a signal state of the balanced signal, and the signal states corresponding to the two state positions are different.

Optionally, the rough judgment module is specifically configured to: determine the dimensionality reduction position information according to the equalized signal; determine the dimensionality reduction signal according to the equalization signal and the dimensionality reduction position information.

Optionally, the equalized signal includes equalized sub-signals at each of the n times. Determining the dimension-reduced position information according to the equalized signal includes: determining a position identification value corresponding to each time point according to the equalized sub-signal at each time point in the n time points. Determining the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information includes: determining the dimensionality reduction subsignal at each moment according to the equalization subsignal at each moment in the n moments and the position identification value corresponding to each moment Signal.

Optionally, the equalization module includes: a first equalization submodule and a second equalization submodule. The first equalization submodule, the second equalization submodule and the rough judgment module are connected in sequence, and the first equalization submodule is also connected with the rough judgment module. The first equalization sub-module is used to perform a first equalization process on the initial signal to obtain a first equalized signal. The second equalization sub-module is used for performing a second equalization process on the first equalized signal to obtain a second equalized signal. Wherein, the equalization signal obtained by the equalization module includes a first equalization signal and a second equalization signal, and the corresponding position of the dimensionality reduction sub-signal in the equalization signal is the corresponding position of the dimensionality reduction sub-signal in the second equalization signal, and the equalization sub-signal is a sub-signal in the second equalized signal. Wherein, the first equalized signal and the second equalized signal respectively include equalized sub-signals at each of the n times.

In the technical solution provided by this application, since the equalization module includes two cascaded equalization sub-modules, the equalization module can perform two-level equalization processing on the initial signal input to the equalization module, which helps to improve the equalization signal output by the equalization module (section The signal-to-noise ratio of the second equalization signal), thereby improving the accuracy of the dimensionality reduction position information determined by the rough decision module and the dimensionality reduction signal.

Optionally, the first equalization sub-module is an FFE or a multiple-input multiple-output (multiple-input multiple-out-put, MIMO) equalizer. The second equalization sub-module is decision feedback equalization (decision feedback equalization, DFE).

Optionally, when the equalization module includes a first equalization submodule and a second equalization submodule, the rough decision module is specifically used to: determine the dimensionality reduction position information according to the second equalization signal; determine the dimensionality reduction position information according to the first equalization signal and the dimensionality reduction position information Dimensionality reduction signal. For example, determining the dimensionality reduction position information by the rough decision module according to the second equalized signal includes: the rough decision module determines the corresponding The location identifier value. The coarse judgment module determines the dimensionality reduction signal according to the first equalization signal and the dimensionality reduction position information includes: the rough judgment module according to the equalization sub-signal at each moment of the n moments included in the first equalization signal and the corresponding Determine the dimensionality reduction sub-signal at each moment.

Optionally, the filtered signal includes a filtered sub-signal at each of the n moments, the target path branch includes a decision value at each of the n moments, and each of the n moments The decision value of is obtained by the MLSE module based on the filtered sub-signals at each moment, and the target signal includes target sub-signals at each of the n times. The mapping module is specifically used to: determine the target sub-signal at each time according to the decision value at each time in the n times and the position identification value corresponding to each time; Signal identifies the target signal.

Optionally, the judgment value is a hard value, and the mapping module is specifically configured to: determine the sum of the judgment value at each of the n moments and the position identification value corresponding to each moment as the target subsignal.

Optionally, the decision value is a soft value, and the mapping module is specifically configured to: determine the mapping rule corresponding to each moment according to the position identification value corresponding to each moment in the n moments; The value and the mapping rule corresponding to each moment determine the target sub-signal at each moment.

Optionally, the signal state of the equalized signal includes -3, -1, 1, and 3, the signal state of the dimensionality reduction signal includes -1 and 1, and the position identification value includes -2, 0, and 2. When p_ind_d(k)=-2, the mapping rule is hd_out(k)=[max, -ld_out(k)].

When p_ind_d(k)=0, the mapping rule is hd_out(k)=[ld_out(k), max]. When p_ind_d(k)=2, the mapping rule is hd_out(k)=[-max, ld_out(k)]. p_ind_d(k) is the position identification value corresponding to the kth moment, ld_out(k) is the judgment value at the kth moment, hd_out(k) is the target sub-signal at the kth moment, max is a constant, k is a positive value not greater than n Integer, in the expression [A, B] of the mapping rule, A is the high-order bit of the target sub-signal, and B is the low-order bit of the target sub-signal.

Optionally, the signal processing device further includes: an autocorrelation estimation module. The autocorrelation estimation module is respectively connected with the equalization module, the filtering module and the MLSE module. The autocorrelation estimation module is used to determine the filter coefficient of the filter module according to the equalized signal, and provide the filter coefficient to the filter module and the MLSE module. The filtering module is specifically configured to perform filtering processing on the dimensionality reduction signal according to the filtering coefficient. The MLSE module is specifically used to determine the target path branch of the filtered signal according to the filter coefficient.

Optionally, the filtered signal includes filtered sub-signals at each of the n times, and the MLSE module is specifically configured to: determine according to the filter coefficient of the filtering module and the filtered sub-signal at each of the n times The branch metrics at each moment; determining the target path branch of the filtered signal according to the branch metrics at n times.

Optionally, the signal processing device further includes: a decoding module. The decoding module is connected with the mapping module. The decoding module is used to decode the target signal to restore data. Wherein, the decoding module may include a forward error correction code (forward error correction, FEC) decoder, and the decoding module 107 recovers the data as a sequence composed of 0 and 1.

Optionally, the signal processing device is an intensity modulation device or a phase modulation device.

Optionally, the equalized signal is a complex signal or a real signal.

Optionally, the equalization signal is a PAM signal, and the dimensionality reduction signal is a binary amplitude keying (on-off keying, OOK) signal. Alternatively, the equalized signal is a quadrature amplitude modulation (quadrature amplitude modulation, QAM) signal, and the dimensionality reduction signal is a quadrature phase shift keying (quadrature phase shift keying, QPSK) signal.

In a second aspect, a signal processing method is provided, which is applied to a signal processing device. The signal processing device includes an equalization module, a rough judgment module, a filter module, an MLSE module and a mapping module connected in sequence, and the rough judgment module is also connected with the mapping module. The method includes: an equalization module performs equalization processing on an initial signal input to the equalization module to obtain an equalized signal. The rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalization signal, and the dimensionality reduction signal includes the dimensionality reduction sub-signals at each moment in n times , the dimensionality reduction signal has nothing to do with the modulation format, the dimensionality reduction position information includes the position identification value corresponding to each of the n times, and the position identification value corresponding to each time point is used to indicate that the dimensionality reduction sub-signal at each time point is at The corresponding position in the equalized signal, n is a positive integer. The filtering module performs filtering processing on the dimensionality reduction signal to obtain the filtered signal. The MLSE module determines target path branches of the filtered signal. The mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal. The status is the same or corresponds.

Optionally, the rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalized signal, including: the rough judgment module determines the dimensionality reduction position information according to the equalized signal; the rough judgment module determines the dimensionality reduction signal according to the equalized signal and the dimensionality reduction position information.

Optionally, the equalized signal includes equalized sub-signals at each of the n moments, and the rough decision module determines the dimensionality reduction position information according to the equalized signal, including: the rough decision module The equalized sub-signal of , and determine the position identification value corresponding to each moment. The rough judgment module determines the dimensionality reduction signal according to the equalized signal and the dimensionality reduction position information, including: the rough judgment module determines each Dimensionality reduction sub-signal at a moment.

Optionally, the equalization module includes a first equalization submodule and a second equalization submodule. The first equalization sub-module, the second equalization sub-module and the rough judgment module are connected in sequence, and the first equalization sub-module is also connected with the rough judgment module. The equalization module equalizes the initial signal input to the equalization module to obtain an equalized signal, including: the first equalization sub-module performs first equalization processing on the initial signal to obtain the first equalized signal; the second equalization sub-module performs equalization on the first equalized signal second equalization processing to obtain a second equalization signal. Wherein, the equalization signal obtained by the equalization module includes a first equalization signal and a second equalization signal, and the corresponding position of the dimensionality reduction sub-signal in the equalization signal is the corresponding position of the dimensionality reduction sub-signal in the second equalization signal, and the equalization sub-signal is a sub-signal in the second equalized signal.

Optionally, the rough decision module determining the dimensionality reduction location information according to the equalized signal includes: the rough decision module determining the dimensionality reduction location information according to the second equalization signal. The rough judgment module determines the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information, including: the rough judgment module determines the dimensionality reduction signal according to the first equalization signal and the dimensionality reduction location information.

Optionally, the filtered signal includes a filtered sub-signal at each of the n moments, the target path branch includes a decision value at each of the n moments, and each of the n moments The decision value of is obtained by the MLSE module based on the filtered sub-signals at each moment, and the target signal includes target sub-signals at each of the n times. The mapping module maps the target path branch of the filtered signal to obtain the target signal according to the dimensionality reduction position information, including: the mapping module according to the decision value at each time in the n times and the position identification value corresponding to each time, Determine the target sub-signal at each moment; the mapping module determines the target signal according to the target sub-signals at n times.

Optionally, the decision value is a hard value, and the mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment, including: mapping The module determines the sum of the decision value at each of the n moments and the position identification value corresponding to each moment as the target sub-signal at each moment.

Optionally, the decision value is a soft value, and the mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment, including: mapping The module determines the mapping rule corresponding to each time according to the position identification value corresponding to each time in the n times; The mapping rule determines the target sub-signal at each moment.

Optionally, the signal state of the equalized signal includes -3, -1, 1, and 3, the signal state of the dimensionality reduction signal includes -1 and 1, and the position identification value includes -2, 0, and 2. When p_ind_d(k)=-2, the mapping rule is hd_out(k)=[max, -ld_out(k)]. When p_ind_d(k)=0, the mapping rule is hd_out(k)=[ld_out(k), max]. When p_ind_d(k)=2, the mapping rule is hd_out(k)=[-max, ld_out(k)]. p_ind_d(k) is the position identification value corresponding to the kth moment, ld_out(k) is the judgment value at the kth moment, hd_out(k) is the target sub-signal at the kth moment, max is a constant, k is a positive value not greater than n Integer, in the expression [A, B] of the mapping rule, A is the high-order bit of the target sub-signal, and B is the low-order bit of the target sub-signal.

Optionally, the signal processing device further includes an autocorrelation estimation module, and the autocorrelation estimation module is respectively connected to the equalization module, the filtering module and the MLSE module. The method also includes: the autocorrelation estimation module determines the filter coefficient of the filter module according to the equalized signal, and provides the filter coefficient to the filter module and the MLSE module respectively. The filtering module performs filtering processing on the dimensionality reduction signal, including: the filtering module performs filtering processing on the dimensionality reduction signal according to the filtering coefficient. The MLSE module determines the target path branch of the filtered signal, including: the MLSE module determines the target path branch of the filtered signal according to the filter coefficient.

Optionally, the filtered signal includes filtered sub-signals at each of the n times, and the MLSE module determines the target path branch of the filtered signal according to the filter coefficient, including: the MLSE module determines the target path branch of the filtered signal according to the filter coefficient and the n times The filtered sub-signal at each moment determines the branch metric at each moment; the MLSE module determines the target path branch of the filtered signal according to the branch metrics at n moments.

Optionally, the signal processing device further includes a decoding module connected to the mapping module. The method also includes: the decoding module decodes the target signal to restore the data.

A third aspect provides a processing chip, where the processing chip includes the signal processing apparatus provided in the first aspect or any optional implementation manner of the first aspect. Optionally, the processing chip is a DSP chip.

In the fourth aspect, a signal receiving device is provided, including a memory chip and a processing chip;

Memory chips are used to store computer programs;

The processing chip is used to execute the computer program stored in the storage chip so that the signal receiving device executes the signal processing method provided in the second aspect or any optional implementation manner of the second aspect.

A fifth aspect provides a signal transmission system, including: a signal sending device and the signal receiving device as provided in the fourth aspect. The signal sending device is connected to the signal receiving device through a transmission link, and the signal sending device is used to send a signal to the signal receiving device through the transmission link.

Optionally, the signal sending device is an optical transmitter, the signal receiving device is an optical receiver, and the transmission link is an optical link.

A sixth aspect provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed, the implementation as provided in the second aspect or any optional implementation manner of the second aspect is provided. signal processing method.

In a seventh aspect, a computer program product is provided, the computer program product includes a program or code, and when the program or code is executed, it realizes the signal processing provided in the second aspect or any optional implementation manner of the second aspect method.

For the technical effects of the above-mentioned second to seventh aspects, reference may be made to the technical effects of the first aspect, which will not be repeated here.

The beneficial effects brought by the technical solution provided by the application are:

In the signal processing method and device, processing chip, and signal transmission system provided in the embodiments of the present application, the equalization module performs equalization processing on the initial signal to obtain an equalized signal, and the rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, and converts the dimensionality reduction The dimension position information is output to the mapping module, and the dimension reduction signal is output to the filter module. The filter module filters the dimension reduction signal to obtain a filter signal, and outputs the filter signal to the MLSE module. The MLSE module determines the target path branch of the filter signal, and The target path branch of the filtered signal is output to the mapping module, and the mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal. Since the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalized signal, and the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, so the filtered signal input to the MLSE module The number of signal states of is less than the number of signal states of the equalized signal output by the equalization module, and the filtered signal input to the MLSE module has nothing to do with the modulation format, so that the MLSE module needs to calculate fewer path branches in the process of determining the target path branch, and The process of determining the branch of the target path by the MLSE module has nothing to do with the modulation format, which helps to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module.

Description of drawings

FIG. 1 is a schematic structural diagram of a signal transmission system provided by an embodiment of the present application;

Fig. 2 is the structural representation of a kind of DSP chip provided by related art;

FIG. 3 is a schematic structural diagram of a signal processing device provided in an embodiment of the present application;

Fig. 4 is a corresponding relationship diagram between a position identification value and a signal state of a PAM4 signal provided by an embodiment of the present application;

FIG. 5 is a relationship diagram between an equalization signal and a dimensionality reduction signal provided by an embodiment of the present application;

FIG. 6 is a relationship diagram between another equalization signal and a dimensionality reduction signal provided by an embodiment of the present application;

FIG. 7 is a schematic structural diagram of another signal processing device provided by an embodiment of the present application;

FIG. 8 is a schematic diagram of a mapping rule provided by an embodiment of the present application;

FIG. 9 is a schematic structural diagram of another signal processing device provided by an embodiment of the present application;

FIG. 10 is a schematic structural diagram of another signal processing device provided by an embodiment of the present application;

Fig. 11 is a schematic structural diagram of another signal processing device provided by an embodiment of the present application;

Fig. 12 is a schematic structural diagram of another signal processing device provided by an embodiment of the present application;

FIG. 13 is a schematic diagram of a signal processing process provided by an embodiment of the present application;

FIG. 14 is a relationship diagram between signal power and BER provided by an embodiment of the present application;

FIG. 15 is a graph showing the relationship between power and BER of another signal provided by the embodiment of the present application;

Fig. 16 is a flowchart of a signal processing method provided by an embodiment of the present application.

Detailed ways

In order to make the principles, technical solutions and advantages of the present application clearer, the implementation manners of the present application will be further described in detail below in conjunction with the accompanying drawings.

The signal transmission system usually includes a signal sending device and a signal receiving device, the signal sending device and the signal receiving device are connected through a transmission link, and the signal sending device sends a signal to the signal receiving device through the transmission link. Wherein, the signal transmission system may be an optical transmission system, correspondingly, the signal sending device may be an optical transmitter, the signal receiving device may be an optical receiver, and the transmission link may be an optical link. Alternatively, the signal transmission system may be a wireless transmission system, correspondingly, the signal sending device may be a wireless sending device, the signal receiving device may be a wireless receiving device, and the transmission link may be a wireless link. The signal transmission system may also be a cable transmission system, and accordingly, the transmission link may be a cable link. The transmission link includes a transmission medium and a transmission device. For example, an optical link usually includes an optical transmission medium such as an optical fiber, and may also include optical devices such as an optical amplifier and an optical connector, which are not limited in this embodiment of the present application.

For example, please refer to FIG. 1 , which shows a schematic structural diagram of a signal transmission system provided by an embodiment of the present application. Figure 1 takes the optical transmission system as an example to illustrate. The signal transmission system includes a signal sending device 01 and a signal receiving device 02 , and the signal sending device 01 and the signal receiving device 02 are connected through a transmission link 03 . Wherein, the signal sending device 01 includes a sequentially connected laser (laser) 011, driver (driver) 012, and modulator (modulator) 013, and the signal receiving device 02 includes a sequentially connected photoelectric conversion device (PD) 021 , an analog-digital converter (analog-digital converter, ADC) 022 and a DSP chip 023. The signal sending device 01 and the signal receiving device 02 are connected through the transmission link 03, specifically, the modulator 013 and the PD 021 are connected through the transmission link 03. The optical signal sent by the laser 011 is driven by the driver 012 and then transmitted to the modulator 013. After the modulator 013 modulates the optical signal with the data to be sent, the modulated optical signal is sent to the signal receiving device 02 through the transmission link 03. (The modulated optical signal carries the data to be sent). After the optical signal is transmitted to the signal receiving device 02, the PD 021 converts the optical signal into an electrical signal, and the electrical signal converted by the PD 021 is an analog electrical signal. The ADC performs analog-to-digital conversion on the analog electrical signal to obtain a digital electrical signal. The digital electrical signal is processed to recover the data.

During the signal transmission process, limited by the bandwidth of each device in the signal transmission system, the signal usually has ISI, which affects the quality of the signal. In order to ensure that the signal receiving device recovers correct data from the received signal, after receiving the signal, the signal receiving device usually uses an equalization algorithm to equalize the signal to eliminate or weaken the ISI of the signal. Current equalization algorithms include FFE algorithm, DFE algorithm, MLSE algorithm, MAP (mapreduce) algorithm, etc. Compared with FFE algorithm, DFE algorithm, and MAP algorithm, the bit error rate (bit error rate; BER) of MLSE algorithm is lower, and The resource consumption is less, so the MLSE algorithm is usually used to equalize the signal. Wherein, the process of equalizing the signal is usually performed by the DSP chip in the signal receiving device.

As an example, please refer to FIG. 2 , which shows a schematic structural diagram of a DSP chip provided in the related art. The DSP chip includes an FFE, a post-filter and an MLSE module connected in sequence, and a path branch calculation module connected to the MLSE module and the post-filter respectively. The FFE performs equalization processing on the received signal (that is, a digital electrical signal) to obtain an equalized signal. The equalized signal has no ISI and is non-linear, but the equalized signal contains colored noise. The post-filter filters the equalized signal to obtain a filtered signal. During the process of filtering the equalized signal, the post-filter converts the colored noise in the equalized signal into white noise, and introduces a variable noise into the equalized signal. controlled ISI. The path branch calculation module calculates the number of path branches of the filtered signal according to the filter coefficients of the post filter. The MLSE module performs MLSE processing on the filtered signal according to the number of path branches of the filtered signal to obtain an optimal solution.

Among them, the complexity of the MLSE algorithm depends on the number of path branches input to the MLSE module. Assume that the number of states of the filtered signal is M, and the modulation order of the post-filter is L (correspondingly, the length of the memory introduced is L-1, and the length of the memory introduced refers to the length of the controllable ISI introduced by the post-filter) , then the number of path branches input to the MLSE module is M ^L . That is to say, the number of path branches input to the MLSE module is positively correlated with the modulation order of the post-filter and the number of states of the filtered signal. The higher the modulation order of the post-filter is, the more the number of states of the filtered signal is. The higher the complexity.

In order to reduce the complexity of the MLSE algorithm and reduce the resource consumption of the MLSE process, a second-order post-filter is usually used in the current DSP chip (that is, the modulation order of the post-filter is L=2, and correspondingly, the memory length is introduced as L-1=1) performs signal filtering. However, when a 2-order post-filter is used, the resource consumption of the MLSE process is still relatively large. For example, for a PAM4 signal, when a second-order post-filter is used in the DSP chip, the number of path branches input to the MLSE module is 16, and the MLSE module performs MLSE processing on the filtered signal output by the post-filter according to the number of 16 path branches. resources up to 1 million logic gates. For PAM16 signals, the algorithm complexity of the MLSE module is higher and the resource consumption is greater.

In view of the problems of high algorithm complexity and large resource consumption of the MLSE module in the related art, embodiments of the present application provide a signal processing method and device, a processing chip, and a signal transmission system. In the technical solution provided by the embodiment of the present application, the signal processing device includes an equalization module, a rough decision module, a filtering module, an MLSE module and a mapping module. The equalization module performs equalization processing on the initial signal to obtain an equalized signal. The rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, outputs the dimensionality reduction position information to the mapping module, and outputs the dimensionality reduction signal to the filtering module. The filtering module filters the dimensionality reduction signal to obtain a filtered signal, and outputs the filtered signal to the MLSE module. The MLSE module determines the target path branch of the filtered signal (eg, the optimal path branch), and outputs the target path branch of the filtered signal to the mapping module. The mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal (for example, the optimal signal). Among them, the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalized signal, and the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, so the input filter of the MLSE module The number of signal states of the signal is less than the number of signal states of the equalized signal output by the equalization module, and the filtered signal input to the MLSE module has nothing to do with the modulation format, so that the MLSE module needs to calculate fewer path branches in the process of determining the target path branch, Moreover, the process of determining the branch of the target path by the MLSE module has nothing to do with the modulation format. In this way, it is helpful to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module. The technical solutions of the embodiments of the present application are introduced below.

Please refer to FIG. 3 , which shows a schematic structural diagram of a signal processing device 10 provided by an embodiment of the present application. The signal processing apparatus 10 may be a signal receiving device, or a functional component in the signal receiving device. For example, the signal processing device 10 is a DSP chip in the signal receiving device; or, the signal processing device 10 is a partial structure in the DSP chip; or, the signal processing device 10 includes a DSP chip and other structures in the signal receiving device, and the present application implements Examples are not limited to this.

As shown in FIG. 3 , the signal processing device 10 includes: an equalization module 101 , a rough decision module 102 , a filter module 103 , an MLSE module 104 and a mapping module 105 connected in sequence, and the rough decision module 102 is also connected to the mapping module 105 . For example, the rough decision module 102 has a first output terminal a1 and a second output terminal a2, the mapping module 105 has a first input terminal b1 and a second input terminal b2, and the first output terminal a1 of the rough decision module 102 is connected to the filter module 103 , the second output terminal a2 of the rough decision module 102 is connected to the second input terminal b2 of the mapping module 105 , and the first input terminal b1 of the mapping module 105 is connected to the MLSE module 104 . The equalization module 101 is configured to perform equalization processing on an initial signal input to the equalization module 101 to obtain an equalized signal. The rough decision module 102 is configured to determine dimensionality reduction position information and dimensionality reduction signals according to the equalized signal. After the rough decision module 102 determines the dimension reduction position information and the dimension reduction signal, the dimension reduction signal can be output through the first output terminal a1 of the rough decision module 102, and the dimension reduction position information can be output through the second output terminal a2 of the rough decision module 102 . The filtering module 103 is configured to perform filtering processing on the dimensionality reduction signal to obtain a filtered signal. The MLSE module 104 is used to determine a target path branch (e.g., an optimal path branch) for the filtered signal. The mapping module 105 is used for the dimensionality reduction position information, and maps the target path branch of the filtered signal to obtain a target signal (eg, an optimal signal).

Wherein, the equalized signal has no ISI and is non-linear, but the equalized signal contains colored noise, so the dimensionality reduction signal contains colored noise, and the filtering module 103 converts the colored noise in the dimensionality reduction signal into is white noise, and introduces controllable ISI into the dimensionality reduction signal, so the filtered signal contains white noise and controllable ISI.

Wherein, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalization signal, the dimensionality reduction signal includes dimensionality reduction sub-signals at each moment in n times, the dimensionality reduction signal has nothing to do with the modulation format, the dimensionality reduction signal The location information is related to the modulation format of the equalized signal (the dimensionality-reduced location information may be determined according to the modulation format of the equalized signal), and the dimensionality-reduced location information includes the location identification value corresponding to each moment in the n times, each The position identification value corresponding to each moment is used to indicate the corresponding position of the dimensionality reduction sub-signal at each moment in the equalized signal, and n is a positive integer.

Among them, the number of signal states of the equalized signal is equal to the number of signal states of the initial signal, the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, and the number of signal states of the target signal is equal to the number of signal states of the equalized signal The numbers are equal, and each signal state of the target signal is the same as or corresponds to each signal state of the equalized signal. For example, the target path branch output by the MLSE module 104 includes decision values at each of the n time moments, and the decision values in the target path branch may be hard values or soft values. If the decision value in this target path branch is a hard value, then the individual signal states of the target signal are identical to the individual signal states of the equalized signal. If the decision value in the target path branch is a soft value, the respective signal states of the target signal correspond to the respective signal states of the equalized signal.

Wherein, the initial signal may be a real number signal, and correspondingly, the equalization signal, the dimensionality reduction signal, the filtering signal and the target signal are all real number signals. Alternatively, the initial signal may be a complex signal, and accordingly, the equalized signal, the dimensionality reduction signal, the filtered signal, and the target signal are all complex signals. A real signal includes only the real part, and a complex signal includes both real and imaginary parts. In the embodiment of the present application, the number of signal states of the dimensionality reduction signal and the number of signal states of the filtered signal are smaller than the number of signal states of the initial signal, the number of signal states of the equalized signal, the number of signal states of the target signal, and the initial Signals have an equal number of signal states. For example, the initial signal, the equalization signal, and the target signal are all real signals, and specifically are PAM signals (such as PAM4 signals, PAM16 signals, etc.), then both the dimensionality reduction signal and the filtered signal can be OOK signals. Alternatively, the initial signal, the equalized signal and the target signal are all complex signals, and specifically are QAM signals (such as 64QAM signals, 16QAM signals, etc.), then both the dimensionality reduction signal and the filtered signal may be QPSK signals.

Optionally, the signal processing device 10 may be an intensity modulation device or a phase modulation device. If the signal processing device 10 is an intensity modulation device, the initial signal, the equalized signal, the dimensionality reduction signal, the filtered signal and the target signal are all real number signals. If the signal processing device 10 is a phase modulation device, the initial signal, the equalized signal, the dimensionality reduction signal, the filtered signal and the target signal are all complex signals.

In summary, in the signal processing device provided by the embodiment of the present application, the equalization module performs equalization processing on the initial signal to obtain an equalized signal, and the rough decision module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, and outputs the dimensionality reduction position information To the mapping module, output the dimension reduction signal to the filter module, the filter module filters the dimension reduction signal to obtain the filter signal, and outputs the filter signal to the MLSE module, the MLSE module determines the target path branch of the filter signal, and the target path of the filter signal The path branch is output to the mapping module, and the mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal. Since the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalized signal, and the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, so the filtered signal input to the MLSE module The number of signal states of is less than the number of signal states of the equalized signal output by the equalization module, and the filtered signal input to the MLSE module has nothing to do with the modulation format, so that the MLSE module needs to calculate fewer path branches in the process of determining the target path branch, and The process of determining the branch of the target path by the MLSE module has nothing to do with the modulation format, which helps to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module.

In the embodiment of the present application, for the corresponding position of the dimensionality reduction sub-signal at each of the n times in the equalized signal, the equalized signal has two state positions adjacent to the corresponding position, and the corresponding position Located between the two state positions, the corresponding position is equal to the Euclidean distance between the two state positions, each state position in the two state positions corresponds to a signal state of the balanced signal, and the two state positions The corresponding signal states are different.

Optionally, the equalized signal is a PAM4 signal, and the signal states of the PAM4 signal include -3, -1, 1 and 3. Assuming that the dimensionality reduction signal includes the dimensionality reduction sub-signal ld_sig(k)=-3 at the kth moment, the position identification value corresponding to the kth moment is p_ind_d(k)=-2, and the position identification value indicates the dimensionality reduction sub-signal ld_sig( The corresponding position of k)=-3 in the equalized signal (assumed to be the corresponding position k), then the equalized signal has two state positions adjacent to the corresponding position k, and the two state positions are consistent with the signal of the equalized signal The states -3 and -1 are in one-to-one correspondence, the corresponding position k is located between the two state positions, and the corresponding Euclidean distance between the two state positions is equal to the corresponding position k. Assume again that the dimensionality reduction signal includes the dimensionality reduction sub-signal ld_sig(k)=-1 at the kth moment, and the position identifier value corresponding to the kth moment is p_ind_d(k)=0, which indicates the dimensionality reduction sub-signal ld_sig( k)=-1 in the corresponding position k in the balanced signal, then the balanced signal has two state positions adjacent to the corresponding position k, and the two state positions are equal to the signal states-1 and 1 of the balanced signal One correspondence, the corresponding position k is located between the two state positions, and the Euclidean distance between the corresponding position k and the two state positions is equal. It is also assumed that the dimensionality reduction signal includes the dimensionality reduction sub-signal ld_sig(k)=3 at the kth moment, and the position identification value corresponding to the kth moment is p_ind_d(k)=2, and the position identification value indicates the dimensionality reduction sub-signal ld_sig(k )=3 corresponding position k in the balanced signal, then the balanced signal has two state positions adjacent to the corresponding position k, and the two state positions correspond one to one to the signal states 1 and 3 of the balanced signal, The corresponding position k is located between the two state positions, and the corresponding position k is equal to the Euclidean distance between the two state positions. Wherein, the kth time is any one of the n times, 1≤k≤n, and k is an integer.

As an example, the equalized signal is a PAM4 signal. Please refer to FIG. 4 , which shows a correspondence diagram between a location identifier value and a signal state of a PAM4 signal provided in an embodiment of the present application. The corresponding position indicated by the position identification value p_ind=-2 (this corresponding position is referred to as corresponding position-2 for brevity) is located in the state position corresponding to signal state-3 (this state position is referred to as state position-3 for brevity) and signal Between the state positions corresponding to state-1 (this state position is referred to as state position-1 for brevity), and the Euclidean distance between the corresponding position-2 and the state position-3 is equal to the distance between the corresponding position-2 and the state position-1 Euclidean distance between . The corresponding position indicated by the position identification value p_ind=0 (this corresponding position is referred to as corresponding position 0 for brevity) is located in the state position corresponding to signal state-1 (this state position is referred to as state position-1 for brevity) and signal state 1 Between the corresponding state positions (this state position is referred to as state position 1 for brevity), and the Euclidean distance between the corresponding position 0 and the state position -1 is equal to the Euclidean distance between the corresponding position 0 and the state position 1. The corresponding position indicated by the position identification value p_ind=2 (for brevity, this corresponding position is referred to as corresponding position 2) is located at the state position corresponding to signal state 1 (for brevity, this state position is referred to as state position 1) corresponding to signal state 3 The Euclidean distance between the state positions (this state position is called state position 3 for brevity), and the Euclidean distance between the corresponding position 2 and the state position 1 is equal to the Euclidean distance between the corresponding position 2 and the state position 3.

In the embodiment of the present application, the rough judgment module 102 is used to determine the dimensionality reduction position information and the dimensionality reduction signal according to the equalized signal specifically includes: the rough judgment module 102 is specifically used to: determine the dimensionality reduction position information according to the equalized signal; according to the equalized signal and The dimensionality reduction position information determines the dimensionality reduction signal. The realization process of determining the dimensionality reduction position information by the rough decision module 102 and the realization process of the dimensionality reduction signal determination by the rough decision module 102 are respectively introduced below.

In an optional implementation of the embodiment of the present application, the equalized signal includes equalized sub-signals at each of the n times, and the rough decision module 102 determines the dimensionality reduction position information according to the equalized signal includes: the rough decision module 102 according to the For the equalized sub-signal at each of the n time instants, determine the position identification value corresponding to each instant. As an example, the rough judgment module 102 judges the balanced sub-signal at each time according to the judgment rule, and determines the position identification value corresponding to each time according to the judgment result of the balanced sub-signal at each time. Optionally, the decision result of the equalized sub-signal at each moment includes two decision signal states (or called decision symbols), and the rough decision module 102 determines the position identification value corresponding to each moment according to the two decision signal states, The two decision signal states correspond to two state positions, and the corresponding position indicated by the position identification value is equal to the Euclidean distance between the two state positions.

Taking the balanced signal as a PAM4 signal as an example, the decision rules can be shown in Table 1 below:

Table 1

均衡子信号equalizer sub-signal	判决结果verdict
hd_sig(k)<-1hd_sig(k)<-1	-3和-1-3 and -1
-1≤hd_sig(k)≤1-1≤hd_sig(k)≤1	-1和1-1 and 1
hd_sig(k)>1hd_sig(k)>1	1和31 and 3

hd_sig(k) represents the equalized sub-signal at the kth moment, and each of -3, -1, 1 and 3 in the decision result is a decision signal state (or called a decision symbol). According to Table 1, when hd_sig(k)<-1, the judgment result is -3 and -1; when -1≤hd_sig(k)≤1, the judgment result is -1 and 1; when hd_sig(k)> 1, the verdict is 1 and 3.

For example, assuming that hd_sig(k)=-3, since hd_sig(k)<-1, the rough decision module 102 determines that the decision result at the kth moment is -3 and -1 according to the decision rule shown in Table 1, and thus according to The decision result determines the position identifier value p_ind_d(k)=[-3+(-1)]/2=-2 corresponding to the kth moment. As another example, assuming that hd_sig(k)=-1, since -1≤hd_sig(k)≤1, the rough decision module 102 determines that the decision result at the kth moment is -1 and 1 according to the decision rule shown in Table 1, Therefore, the position identifier value p_ind_d(k)=(-1+1)/2=0 corresponding to the kth moment is determined according to the decision result. As another example, assuming that hd_sig(k)=3, since hd_sig(k)>1, the rough judgment module 102 determines that the judgment results at the kth moment are 1 and 3 according to the judgment rules shown in Table 1, so that according to the judgment results It is determined that p_ind_d(k)=(1+3)/2=2. Wherein, the relationship between the decision result and the location identification value may be as shown in FIG. 4 .

In an optional implementation of the embodiment of the present application, the equalized signal includes equalized sub-signals at each of the n times, and the rough judgment module 102 determines the dimensionality reduction signal according to the equalization signal and dimensionality reduction position information including: rough judgment The module 102 determines the dimensionality reduction sub-signal at each time point according to the balanced sub-signal at each time point in the n time points and the location identification value corresponding to each time point. As an example, the rough decision module 102 uses the dimensionality reduction formula ld_sig(k)=hd_sig(k)-p_ind_d(k) to determine the dimensionality reduction sub-signal at the kth moment, that is, the dimensionality reduction subsignal at the kth moment is equal to the dimensionality reduction subsignal at the kth moment The difference between the equalization sub-signal at time k and the position identification value corresponding to time k. For example, assuming hd_sig(k)=-3, p_ind_d(k)=-2, then ld_sig(k)=-3-(-2)=-1. Suppose hd_sig(k)=-1, p_ind_d(k)=0, then ld_sig(k)=-1-0=-1. Suppose hd_sig(k)=3, p_ind_d(k)=2, then ld_sig(k)=3-2=1.

In the embodiment of the present application, the rough decision module 102 implements dimensionality reduction of the equalized signal, so that the number of signal states of the dimensionality-reduced signal determined by the rough decision module 102 is smaller than the number of signal states of the equalized signal. The signal state refers to the possible states of the signal, and the number of signal states refers to the number of possible states of the signal at each moment. In some embodiments, signal states are also referred to as levels, therefore, the number of signal states is also referred to as the number of signal levels.

Taking the equalized signal as an example of a PAM4 signal, please refer to FIG. 5 , which shows a relationship diagram between an equalized signal and a dimensionality reduction signal provided by an embodiment of the present application. It can be seen from Figure 5 that the signal states of the balanced signal include -3, -1, 1 and 3, the signal states of the dimensionality reduction signal include -1 and 1, the balance signal has 4 signal states, and the dimensionality reduction signal has 2 Signal states, so the number of signal states of the reduced-dimensional signal is smaller than the number of signal states of the equalized signal, and the reduced-dimensional signal is independent of the modulation format of the equalized signal.

In the embodiment of the present application, through the processing of the rough judgment module 102, the most likely signal state (most likely to be the state of the signal sent by the signal sending device) is selected from the signal state of the equalized sub-signal at each moment, and based on this The most likely signal state obtains the signal state of the dimensionality reduction sub-signal at each moment, and the subsequent filtering module 103, MLSE module 104 and other modules process according to the dimensionality reduction sub-signal, which helps to simplify the filtering module 103, MLSE module 104, etc. Computational complexity of the module. Still taking the example that the equalized signal is a PAM4 signal, please refer to FIG. 6 , which shows a relationship diagram between another equalized signal and a dimensionality reduction signal provided by an embodiment of the present application. As shown in Figure 6, the equalized signal includes the equalized sub-signals at the kth moment, the k+1th moment, and the k+2th moment, and the kth moment, the k+1st moment, and the k+2th moment The signal states of the balanced sub-signal at each moment include -3, -1, 1, and 3, and the most likely signal states of the signal states of the balanced sub-signal at the kth moment are -3 and -1, and the k+1th The most probable signal states of the signal states of the equalized sub-signal at time instant are 1 and 3, and the most probable signal states of the signal states of the equalized sub-signal at time k+2 are -1 and 1. The dimensionality reduction signal and dimensionality reduction position information are obtained through the processing of the rough decision module 102 . The dimensionality reduction signal includes the dimensionality reduction sub-signals at the kth moment, the k+1th moment and the k+2th moment, and the dimensionality reduction position information includes the kth moment, the k+1th moment and the k+2th moment The location identifier values corresponding to each moment in . The signal state of the sub-signal at each of the kth time, the k+1 time and the k+2 time includes -1 and 1, the position identification value corresponding to the k time is -2, and the k+1 time corresponds to The position identification value of is 2, and the position identification value corresponding to the k+2th moment is 0. Among them, the position identifier value -2 corresponding to the kth moment is determined according to the most likely signal states -3 and -1 of the equalized sub-signal at the kth moment, and the position identifier value 2 corresponding to the k+1th moment is determined according to the k+1th The most probable signal states 1 and 3 of the balanced sub-signal at the moment are determined, and the position identification value 0 corresponding to the k+2th time is determined according to the most probable signal states -1 and 1 of the balanced sub-signal at the k+2th time .

In the embodiment of the present application, the equalization module 101 may include at least one equalization sub-module. As an example, please refer to FIG. 7 , which shows a schematic structural diagram of another signal processing apparatus 10 provided by an embodiment of the present application. The equalization module 101 includes a first equalization sub-module 1011 and a second equalization sub-module 1012 . The first equalization sub-module 1011 , the second equalization sub-module 1012 and the rough decision module 102 are connected in sequence, and the first equalization sub-module 1011 is also connected to the rough decision module 102 . For example, the rough decision module 102 has a first input terminal a3 and a second input terminal a4, the output terminal of the second equalization sub-module 1012 is connected to the first input terminal a3 of the rough decision module 102, and the output terminal of the first equalization sub-module 1011 It is connected with the second input terminal a4 of the rough decision module 102 . Wherein, the equalization module 101 is used to equalize the initial signal input to the equalization module 101, including: the first equalization sub-module 1011 is used to perform the first equalization process on the initial signal to obtain the first equalized signal (marked as 1hd_sig); The equalization sub-module 1012 is configured to perform a second equalization process on the first equalized signal to obtain a second equalized signal (marked as 2hd_sig). The equalized signal (i.e. hd_sig) obtained by equalizing the initial signal by the equalization module 101 includes a first equalized signal (i.e. 1hd_sig) and a second equalized signal (i.e. 2hd_sig), and the corresponding position of the dimensionality reduction sub-signal in the equalized signal is the Corresponding positions of the dimensionality reduction sub-signals in the second equalized signal 2hd_sig, the first equalized signal and the second equalized signal respectively include the equalized sub-signals at each of the n times.

Wherein, the first equalization sub-module 1011 may be an FFE or a MIMO equalizer, and the second equalization sub-module 1012 may be a DFE. For example, the initial signal is a real number signal, and the first equalization sub-module 1011 is an FFE. For another example, the initial signal is a complex signal, and the first equalization sub-module 1011 is a MIMO equalizer. In the embodiment of the present application, the equalization module 101 includes two sub-modules as an example. In practical applications, the equalization module 101 may include three or more sub-modules, and the sub-modules in the equalization module 101 may be FFE, DFE or MAP equalizers. device, which is not limited in the embodiment of this application. Since the equalization module 101 includes two cascaded equalization sub-modules, the equalization module 101 can carry out two-stage equalization processing on the initial signal input to the equalization module 101, which helps to improve the equalized signal (the second equalized signal) output by the equalized module 101. ), thereby improving the accuracy of the dimensionality reduction position information and the dimensionality reduction signal output by the rough decision module 102.

In the embodiment of the present application, for the signal processing device 10 shown in FIG. 7 , the rough decision module 102 determining the dimensionality reduction position information according to the equalized signal includes: the rough decision module 102 determining the dimensionality reduction position information according to the second equalized signal. For example, the rough decision module 102 determines the position identification value corresponding to each time point according to the equalized sub-signal at each time point in the n time points included in the second equalized signal. The rough decision module 102 determining the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information includes: the rough decision module 102 determining the dimensionality reduction signal according to the first equalization signal and the dimensionality reduction location information. For example, the rough decision module 102 determines the dimensionality reduction sub-signal at each time according to the equalization sub-signal at each time in the n time included in the first equalization signal and the position identification value corresponding to each time. Among them, the rough judgment module 102 determines the process of the position identification value corresponding to each moment according to the equalized sub-signal at each moment included in the second equalized signal, and the rough decision module 102 determines the corresponding position identification value according to each moment included in the first equalized signal The process of determining the dimensionality reduction sub-signal at each moment by the equalization sub-signal and the position identification value corresponding to each moment can refer to the relevant description of the embodiment shown in FIG. The difference between the embodiments is that, for the signal processing device 10 shown in FIG. 7 , the equalized sub-signal used when the rough judgment module 102 determines the position identification value p_ind_d(k) corresponding to the k-th moment is the k-th equalized sub-signal included in the second equalized signal The equalized sub-signal 2hd_sig(k) at the moment, the equalized sub-signal adopted when the rough decision module 102 determines the dimensionality reduction sub-signal ld_sig(k) at the kth moment is the equalized sub-signal 1hd_sig( k). In the signal processing device 10 shown in FIG. 7 , for the corresponding position of the dimensionality reduction sub-signal at each of the n times in the second equalized signal, the second equalized signal has a Two state positions, the corresponding position is located between the two state positions, the corresponding position is equal to the Euclidean distance between the two state positions, and each state position in the two state positions corresponds to the second equalization signal A signal state, the signal states corresponding to the two state positions are different.

In the embodiment of the present application, the filtering module 103 has a filtering coefficient, and the filtering module 103 is configured to perform filtering processing on the dimensionality reduction signal output by the rough decision module 102 according to the filtering coefficient to obtain a filtered signal. Wherein, the filtered signal includes filtered sub-signals at each of the n times. Optionally, the filtering module 103 is a post filter.

In an optional implementation, the filtering module 103 is configured to perform filtering processing on the dimensionality reduction signal output by the rough decision module 102 according to the filtering coefficient, including: the filtering module 103 according to the filtering coefficient c of the filtering module 103, the The dimensionality reduction sub-signal ld_sig(k) and the dimensionality reduction sub-signal ld_sig(k) at the k+1th moment, use the filter formula pf(k+1)=ld_sig(k+1)+c×ld_sig(k) to determine the kth The filtered sub-signal at time +1, pf(k+1) represents the filtered sub-signal at time k+1.

In the embodiment of the present application, the MLSE module 104 is used to determine a target path branch of the filtered signal output by the filtering module 103. In some embodiments, the target path branch is also called an optimal path branch or an optimal sequence. The filtered signal includes filtered sub-signals at each of the n times. In an optional implementation, the MLSE module 104 is specifically configured to: according to the filter coefficient of the filter module 103 and each of the n times The branch metrics at each moment are determined for the filtered sub-signals at n moments; and the target path branch of the filtered signal is determined according to the branch metrics at n moments.

As an optional implementation, the MLSE module 104 determines the codebook (code book) corresponding to the filtered signal according to the filter coefficient of the filter module 103 and the signal state of the filtered signal, and according to the filtered sub-signal at each moment and the filtered signal The corresponding codebook determines the branch metrics at each moment, and the MLSE module 104 can determine multiple branch metrics at each moment, and the MLSE module 104 can determine the target branch metric from the multiple branch metrics at each moment (for example, the most optimal branch metric), and determine the target path branch of the filtered signal according to the target branch metric at n moments. Exemplarily, assuming that the number of branch metrics at each of the n moments is m, the MLSE module 104 determines a target branch metric from the m branch metrics at each moment to obtain n target branch metrics at n moments , the MLSE module 104 determines the target path branch of the filtered signal according to the n target branch metrics.

As another optional implementation, the MLSE module 104 determines the codebook corresponding to the filtered signal according to the filter coefficient of the filtering module 103 and the signal state of the filtered signal, and according to the filtered sub-signal The codebook corresponding to the filtered signal determines the branch metrics at each moment, assuming that the number of branch metrics at each moment is m, the MLSE module 104 can determine m×n branch metrics at the n moments, and the MLSE module 104 Combining and accumulating the m×n branch metrics at n times (a branch metric at each moment is accumulated with all branch metrics at all other times), and determining the target path branch according to the accumulation result. In the embodiment of the present application, the MLSE module 104 may include a branch metric calculation submodule at n time, a cumulative branch metric calculation submodule and an optimal path selection submodule (these submodules are not shown in FIG. 4 and FIG. 7 ), The branch metrics at each of the n moments can be determined by the branch metric calculation submodule at n moments, the branch metrics at the n moments are combined and accumulated by the cumulative branch metric calculation submodule, and determined by the optimal path selection submodule target path branch.

As an example of the embodiment of the present application, the initial signal is a PAM4 signal, the filtered signal is an OOK signal, the signal state of the filtered signal includes -1 and 1, and the MLSE module 104 determines the filter coefficient c according to the signal state of the filtered signal. The codebook corresponding to the signal includes -1+c, -1-c, 1+c and 1-c. The MLSE module 104 determines the branch metric at each moment according to the filtered sub-signal at each moment and the codebook corresponding to the filtered signal includes: the MLSE module 104 adopts the branch metric formula metric(k _p )=(pf(k)-cb _p ) ² Determine the branch metric at the kth moment. Among them, metric(k _p ) represents the p-th branch metric at the k-th moment, cb _p represents the p-th codebook in the codebook corresponding to the filtered signal, and cb _p is -1+c, -1-c, 1 One of +c and 1-c, pf(k) represents the filtered signal at the kth moment. For the filtered signal at each moment, the MLSE module 104 calculates the filtered signal at each moment with each codebook in -1+c, -1-c, 1+c and 1-c, so the MLSE module 104 may obtain the 4 branch metrics at each moment, and the MLSE module 104 may determine the minimum branch metric among the 4 branch metrics at each moment as the target branch metric at each moment.

In the embodiment of the present application, assuming that the number of signal states of the dimensionality reduction signal is M, and the order of the filtering module 103 is L, the algorithm complexity of the MLSE module 104 may be M ^L . For the case where the equalized signal is a PAM4 signal, the dimensionality reduction signal may be an OOK signal, and the algorithm complexity of the MLSE module 104 may be 2 ^L .

In this embodiment of the present application, the target path branch includes decision values at each of the n times, and the decision value at each moment is obtained by the MLSE module 104 based on the filtered sub-signal at each moment. For example, the decision value at each moment may be obtained by making a decision according to the target branch metric at each moment. Wherein, the decision value in the target path branch can be a hard value or a soft value, and the hard value (that is, the decision value) at each moment is -1 or 1, indicating that the MLSE module 104 judges that the signal at each moment is -1 or 1, the target path branch is a target sequence of length n consisting of -1 and 1. The soft value is the probability ratio, also known as the logarithmic likelihood ratio (likelihood rate, LLR) value, the soft value (that is, the decision value) at each moment is the probability that the signal at each moment is 1 and the signal at each moment The ratio of the probabilities of being -1 (the numerator is the probability that the signal is 1, and the denominator is the probability that the signal is -1). The algorithm corresponding to the hard value is the viterbi algorithm, that is, when the MLSE module 104 adopts the viterbi algorithm, the decision value in the target path branch output by the MLSE module 104 is a hard value. The algorithm corresponding to the soft value can be BCJR (Bahl, Cocke Jelinek and Raviv) algorithm or soft output viterbi algorithm (soft output viterbi algorithm, SOVA), that is, in the target path branch of MLSE module 104 output when MLSE module 104 adopts BCJR algorithm or SOVA The judgment value of is a soft value.

In the embodiment of the present application, the target path branch output by the MLSE module 104 (that is, the target path branch of the filtered signal) includes the decision values at each of the n moments, and the mapping module 105 performs a process of filtering the filtered signal according to the dimensionality reduction position information The target path branch of is mapped to obtain the target signal, and the target signal includes target sub-signals at each of the n time points. The mapping module 105 is specifically used to: determine the target sub-signal at each time according to the decision value at each time in the n times and the position identification value corresponding to each time; Signal identifies the target signal. For example, the mapping module 105 determines the target sub-signals at n times as a whole as the target signal.

As mentioned above, the decision value in the target path branch can be hard value or soft value. In the embodiment of the present application, according to the type of the decision value (hard value or soft value) in the target path branch, the mapping module 105 maps the target path branch of the filtered signal according to the dimensionality reduction position information in different ways. The implementation manners of mapping the target path branch of the filtered signal by the mapping module 105 are respectively introduced below by taking the decision value of the target path branch as a hard value or a soft value.

In an optional implementation manner, the decision value in the target path branch is a hard value, and the mapping module 105 is specifically configured to: identify the decision value at each moment in the n moments with the position corresponding to each moment The sum of values is determined as the target sub-signal at each moment. For example, the mapping module 105 uses the mapping formula hd_out(k)=ld_out(k)+p_ind_d(k) to determine the target sub-signal at the kth moment. hd_out(k) represents the target sub-signal at the kth moment, ld_out(k) represents the decision value at the kth moment, and p_ind_d(k) represents the position identification value corresponding to the kth moment.

As an example, assuming ld_out(k)=-1, p_ind_d(k)=-2, then hd_out(k)=-1+(-2)=-3. Suppose ld_out(k)=-1, p_ind_d(k)=0, then hd_out(k)=-1+0=-1. Suppose ld_out(k)=1, p_ind_d(k)=0, then hd_out(k)=1+0=1. Suppose ld_out(k)=1, p_ind_d(k)=2, then hd_out(k)=1+2=3.

In another optional implementation manner, the decision value in the target path branch is a soft value, and the mapping module 105 is specifically configured to: determine the value at each time point according to the position identification value corresponding to each time point in the n time points A corresponding mapping rule: determine the target sub-signal at each time according to the decision value at each time and the corresponding mapping rule at each time.

Taking the balanced signal as a PAM4 signal as an example, the dimensionality reduction signal can be an OOK signal. The signal states of the balanced signal include -3, -1, 1, and 3, the signal states of the dimensionality reduction signal include -1 and 1, and the location identifier values include - 2, 0 and 2. When p_ind_d(k)=-2, the mapping rule may be hd_out(k)=[max, -ld_out(k)]. When p_ind_d(k)=0, the mapping rule may be hd_out(k)=[ld_out(k), max]. When p_ind_d(k)=2, the mapping rule may be hd_out(k)=[-max, ld_out(k)]. Among them, p_ind_d(k) is the position identification value corresponding to the kth moment, ld_out(k) is the judgment value at the kth moment, hd_out(k) is the target sub-signal at the kth moment, max is a constant, and k is not greater than n A positive integer of , in the expression [A, B] of the mapping rule, A is the high-order bit of the target sub-signal, and B is the low-order bit of the target sub-signal. When the judgment value is a soft value, the mapping rule may also be as shown in FIG. 8 .

Exemplarily, assuming that the position identifier value p_ind_d(k)=-2 corresponding to the kth moment, the mapping module 105 determines that the mapping rule corresponding to the kth moment is hd_out(k)=[max,-ld_out(k)], and the mapping module 105 Determine the target sub-signal at the kth moment according to the decision value at the kth moment and the corresponding mapping rule hd_out(k)=[max, -ld_out(k)] at the kth moment. Assuming that the position identification value p_ind_d(k)=0 corresponding to the kth moment, the mapping module 105 determines that the mapping rule corresponding to the kth moment is hd_out(k)=[ld_out(k), max], and the mapping module 105 according to the kth moment The decision value of and the corresponding mapping rule hd_out(k)=[ld_out(k), max] at the kth moment determine the target sub-signal at the kth moment. Assuming that the position identifier value p_ind_d(k)=2 corresponding to the kth moment, the mapping module 105 determines that the mapping rule corresponding to the kth moment is hd_out(k)=[-max, ld_out(k)], and the mapping module 105 according to the kth The decision value at time and the corresponding mapping rule hd_out(k)=[-max, ld_out(k)] at time k to determine the target sub-signal at time k.

According to the introduction of the above two mapping methods, it can be seen that when the decision value in the target path branch is a hard value, each signal state of the target signal is the same as that of the equalized signal, and when the decision value in the target path branch is a soft value , each signal state of the target signal corresponds to each signal state of the equalized signal.

As an optional implementation of the embodiment of the present application, FIG. 9 is a schematic structural diagram of another signal processing device 10 provided in the embodiment of the present application, and FIG. 10 is a schematic structural diagram of another signal processing device 10 provided in the embodiment of the present application. . As shown in FIG. 9 and FIG. 10 , the signal processing device 10 further includes an autocorrelation estimation module 106 . The autocorrelation estimation module 106 is connected to the equalization module 101 , the filtering module 103 and the MLSE module 104 respectively. For example, the input terminal of the autocorrelation estimation module 106 is connected to the output terminal of the equalization module 101 , and the output terminal of the autocorrelation estimation module 106 is respectively connected to the input terminal of the filtering module 103 and the input terminal of the MLSE module 104 . For the signal processing device 10 shown in Figure 9, the input end of the autocorrelation estimation module 106 is connected to the output end of the equalization module 101, and for the signal processing device 10 shown in Figure 10, the input end of the autocorrelation estimation module 106 is connected to the first The output end of the equalization sub-module 1011 is connected. Among them, the autocorrelation estimation module 106 is used to determine the filter coefficient of the filter module 103 according to the equalized signal, and provide the filter coefficient to the filter module 103 and the MLSE module 104 respectively, so that the filter module 103 performs filtering processing on the dimensionality reduction signal according to the filter coefficient , and, the MLSE module 104 determines the target path branch of the filtered signal according to the filter coefficient. Optionally, the autocorrelation estimation module 106 calculates the filter coefficient of the filtering module 103 according to the signal characteristics of the equalized signal (specifically, the first equalized signal output by the first equalization sub-module 1011 in the signal processing device 10 shown in FIG. 10 ). The noise in the equalized signal (specifically the first equalized signal output by the first equalized sub-module 1011 in the signal processing device 10 shown in FIG. 10 ) is colored noise, and the autocorrelation estimation module 106 can calculate the correlation of the colored noise, according to The correlation of colored noise determines the filter coefficients.

As an optional implementation of the embodiment of the present application, Fig. 11 is a schematic structural diagram of another signal processing device 10 provided in the embodiment of the present application, and Fig. 12 is a schematic structural diagram of another signal processing device 10 provided in the embodiment of the present application . As shown in FIG. 11 and FIG. 12 , the signal processing device 10 further includes a decoding module 107 . The decoding module 107 is connected to the mapping module 105 , for example, an input terminal of the decoding module 107 is connected to an output terminal of the mapping module 105 . The decoding module 107 is used for decoding the target signal output by the mapping module 105 to restore the data. Wherein, the decoding module 107 may include an FEC decoder, and the decoding module 107 may also include other decoders, which is not limited in this embodiment of the present application.

Wherein, the data recovered by the decoding module 107 may be a sequence composed of 0 and 1. For example, the signal state of the target signal output by the mapping module 105 includes -3, -1, 1 and 3, as shown in Figure 4 and Figure 8, the bit symbol corresponding to -3 is 00, and the bit symbol corresponding to -1 is 01 , the bit symbol corresponding to 1 is 11, the bit symbol corresponding to 3 is 10, assuming that the target signal output by the mapping module 105 is 3, -1, 1, 3, -3, -1 in turn, then the decoding module 107 according to the target signal The corresponding relationship between the signal state and the bit symbol, the recovered data is 100111100001.

The foregoing embodiments are described by taking the initial signal as a real number signal as an example, and the initial signal may also be a complex number signal. If the initial signal is a complex signal, then the equalized signal, the dimensionality reduction signal, the filtered signal, etc. are all complex signals. For the situation that the initial signal is a complex signal, the position identification value corresponding to each moment is a complex number. The real part of the position identification value corresponding to each time instant, and the imaginary part of the position identification value corresponding to each time instant is determined according to the imaginary part of the equalized sub-signal at each time instant. When the rough judgment module 102 determines the dimensionality reduction sub-signal at each moment, it determines the dimensionality reduction sub-signal at each moment according to the real part of the equalization sub-signal at each moment and the real part of the position identification value corresponding to each moment. For the real part of the signal, the imaginary part of the dimensionality reduction sub-signal at each time is determined according to the imaginary part of the balanced sub-signal at each time and the imaginary part of the position identification value corresponding to each time. The rough judgment module 102 determines the realization process of the real part of the position identification value corresponding to each moment, the rough judgment module 102 determines the realization process of the imaginary part of the position identification value corresponding to each moment, and the rough judgment module 102 determines the realization process of the imaginary part of the position identification value corresponding to each moment. The realization process of the real part of the dimensionality reduction sub-signal, and the realization process of the imaginary part of the dimensionality reduction sub-signal determined by the rough decision module 102 at each moment, can refer to the situation that the initial signal is a real number signal, and will not be repeated here .

For example, please refer to FIG. 13, which shows a schematic diagram of a signal processing process provided by the embodiment of the present application. In FIG. 13, the signal processing device is the signal processing device 10 shown in FIG. 4, and the initial signal is a complex signal As an example. As shown in FIG. 13 , the equalization module 101 equalizes the initial signal input to the equalization module 101 to obtain an equalized signal hd_sig. The equalized signal hd_sig is a complex signal, and the number of signal states of the equalized signal hd_sig is 36. The equalization module 101 outputs the equalized signal hd_sig to the coarse decision module 102. The rough judgment module 102 determines the dimensionality reduction position information p_ind_d and the dimensionality reduction signal ld_sig according to the equalization signal hd_sig, the dimensionality reduction position information p_ind_d is a complex number, the dimensionality reduction signal ld_sig is a complex signal, and the number of signal states of the dimensionality reduction signal ld_sig is 4. The rough decision module 102 outputs the dimensionality reduction signal ld_sig to the filtering module 103 through the first output terminal a1, and outputs the dimensionality reduction position information p_ind_d to the mapping module 105 through the second output terminal a2.

The filtering module 103 performs filtering processing on the dimensionality reduction signal ld_sig to obtain a filtered signal, and outputs the filtered signal to the MLSE module 104 . Although the filtered signal is not shown in FIG. 13 , it can be understood that the filtered signal is a complex signal, and the number of signal states of the filtered signal is four. The MLSE module 104 determines the target path branch ld_out of the filtered signal, and outputs the target path branch ld_out of the filtered signal to the mapping module 105, the target path branch ld_out is a complex signal, and the number of signal states of the target path branch ld_out is four. The mapping module 105 maps the target path branch of the filtered signal according to the dimensionality reduction position information p_ind_d to obtain the target signal hd_out, the target signal hd_out is a complex signal, and the number of signal states of the target signal hd_out is 36. In Figure 13. Small black dots indicate the signal state, and larger dots indicate the most likely signal state.

The signal power and BER in the technical solutions provided by the embodiments of the present application will be described below with reference to the accompanying drawings.

For example, please refer to FIG. 14 , which shows a relationship diagram between signal power and BER provided by an embodiment of the present application. In FIG. 14 , in the signal processing device corresponding to each of

curves

1, 2, 3, 4 and 5, the equalization module 101 only includes FFE. Curve 1 is a relationship diagram between the power of the equalized signal output by the FFE and the BER. Curve 2 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the viterbi algorithm and the signal processing device does not include a rough decision module. Curve 3 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the viterbi algorithm and the signal processing device includes a rough decision module (such as the power and BER of the signal output by the MLSE module 104 in Fig. 4, Fig. 9 or Fig. 11 BER diagram). Curve 4 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the BCIJ algorithm and the signal processing device does not include a rough decision module. Curve 5 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the BCIJ algorithm and the signal processing device includes a coarse decision module (such as the power and BER of the signal output by the MLSE module 104 in Fig. 4, Fig. 9 or Fig. 11 BER diagram). It can be seen from Figure 14 that the trend of curve 2 is the same as that of curve 3, and the distance between curve 2 and curve 3 is small, the trend of curve 4 is the same as that of curve 5, and the distance between curve 4 and curve 5 is the same. The distance between is small, so it can be determined that the performance of the signal output by the MLSE module based on the dimensionality reduction signal is comparable to the signal performance of the signal output by the MLSE module when the non-rough decision module is included in the signal processing device.

As an example, please refer to FIG. 15 , which shows a relationship diagram between power and BER of another signal provided by an embodiment of the present application. In FIG. 15 , the equalization module 101 of the signal processing device corresponding to curve 1 , curve 2 and curve 4 includes only FFE, and the equalization module 101 of the signal processing device corresponding to curve 3 includes FFE and DFE. Curve 1 is a relationship diagram between the power of the equalized signal output by the FFE and the BER. Curve 2 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the viterbi algorithm and the signal processing device includes a rough decision module (such as the power and BER of the signal output by the MLSE module 104 in Fig. 4, Fig. 9 or Fig. 11 BER diagram). Curve 3 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the MLSE module adopts the viterbi algorithm and the signal processing device includes a rough decision module (such as the power and BER of the signal output by the MLSE module 104 in Fig. 7, Fig. 10 or Fig. 12 BER diagram). Curve 4 is a relationship diagram between the power of the signal output by the MLSE module and the BER when the viterbi algorithm adopted by the MLSE module is not included in the signal processing device. It can be seen from Fig. 15 that the trend of curve 3 is the same as that of curve 4, and the distance between curve 3 and curve 4 is small, so it can be determined that: when the equalization module 101 includes FFE and DFE, the MLSE module is based on the reduction The performance of the signal output by the dimensional signal is equivalent to the signal performance of the signal output by the MLSE module when the signal processing device includes the non-coarse judgment module. The trend of curve 2 is quite different from the trend of curve 3, so it can be determined that when the equalization module 101 includes FFE and DFE, the signal output by the MLSE module based on the dimensionality reduction signal is higher than that when the equalization module 101 only includes FFE, and the MLSE module is based on dimensionality reduction The performance of the signal output signal is good.

In the embodiment of the present application, the signal processing apparatus 10 shown in FIG. 7 , FIG. 10 and FIG. 12 may be applied in a scenario with a large bandwidth limit. Using the signal processing device 10 in FIG. 7 , FIG. 10 and FIG. 12 in a large band-limit scenario can ensure the accuracy of the decision result of the rough decision module 102 when the signal state of the low-dimensional signal output by the rough decision module 102 is small.

In summary, in the signal processing device provided by the embodiment of the present application, the equalization module performs equalization processing on the initial signal to obtain an equalized signal, and the rough decision module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, and outputs the dimensionality reduction position information To the mapping module, output the dimension reduction signal to the filter module, the filter module filters the dimension reduction signal to obtain the filter signal, and outputs the filter signal to the MLSE module, the MLSE module determines the target path branch of the filter signal, and the target path of the filter signal The path branch is output to the mapping module, and the mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal. Since the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalized signal, and the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, so the filtered signal input to the MLSE module The number of signal states of is less than the number of signal states of the equalized signal output by the equalization module, and the filtered signal has nothing to do with the modulation format, so that the MLSE module needs to calculate fewer path branches in the process of determining the target path branch, and the MLSE module determines the target path branch The path branching process has nothing to do with the modulation format, which helps to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module.

In the signal processing device provided by the embodiment of the present application, the signals processed by the filtering module and the MLSE module are both dimension-reduced signals, the number of signal states of the dimension-reduced signal is small, and the bit width of the signal is low, which helps Reduce resource consumption of filtering module and MLSE module. In addition, the signal after dimensionality reduction can be the signal of the lowest dimension (such as OOK signal or QPSK signal), so the processing process of the filtering module and the processing process of the MLSE module have nothing to do with the modulation format, further reducing the resource consumption of the filtering module and MLSE module . It is determined through experiments that, compared with the traditional technical solution, the resource consumption of the MLSE module in the signal processing device provided by the embodiment of the present application can be reduced by 30%, and the circuit area can be reduced by 50%. The signal processing device is an intensity modulation device or a phase modulation device, but is not limited thereto. The higher the dimension of the initial signal (the larger the number of signal states), the more obvious the effect of the signal processing device in reducing resource consumption.

The above is the introduction to the embodiment of the device of the present application, and the following describes the embodiment of the method of the present application.

For example, please refer to FIG. 16 , which shows a flowchart of a signal processing method provided by an embodiment of the present application. The signal processing method may be applied to the signal processing apparatus provided in the foregoing embodiments. The signal processing device includes an equalization module, a rough judgment module, a filter module, an MLSE module and a mapping module connected in sequence, and the rough judgment module is also connected with the mapping module. As shown in Figure 16, the method includes:

S1601. The equalization module performs equalization processing on the initial signal input to the equalization module to obtain an equalized signal.

The initial signal is a real signal or a complex signal, and correspondingly, the equalized signal is a real signal or a complex signal. For example, if the initial signal is a PAM signal, then the equalized signal is a PAM signal. For another example, if the initial signal is a QAM signal, then the equalized signal is a QAM signal. The equalized signal has no ISI and is non-linear, but the equalized signal contains colored noise.

Optionally, the equalization module includes a first equalization submodule and a second equalization submodule, the first equalization submodule, the second equalization submodule and the rough decision module are connected in sequence, and the first equalization submodule is also connected with the rough decision module. The equalization module equalizes the initial signal input to the equalization module to obtain an equalized signal, including: the first equalization sub-module performs first equalization processing on the initial signal to obtain the first equalized signal; the second equalization sub-module performs equalization on the first equalized signal second equalization processing to obtain a second equalization signal. Wherein, the equalized signals output by the equalized module are the first equalized signal and the second equalized signal, and the first equalized signal and the second equalized signal respectively include equalized sub-signals at each of the n times.

S1602. The rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the balanced signal, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the balanced signal, and the dimensionality reduction signal includes the dimensionality reduction at each moment in the n moments The sub-signal, the dimension reduction signal has nothing to do with the modulation format, the dimension reduction position information includes the position identification value corresponding to each time in the n times, and the position identification value corresponding to each time is used to indicate the dimension reduction sub-signal at each time corresponding position in the equalized signal.

Wherein, when the equalized signal is a real signal, the dimensionality reduction signal is a real signal. When the equalized signal is a complex signal, the dimensionality-reduced signal is a complex signal. For example, the equalization signal is a PAM signal, and the dimensionality reduction signal is an OOK signal. For another example, the equalized signal is a QAM signal, and the dimensionality reduction signal is a QPSK signal.

Wherein, for the corresponding position of the dimensionality reduction sub-signal at each time in the n times in the balanced signal, the balanced signal has two state positions adjacent to the corresponding position, and the corresponding position is located in the two states Between positions, the Euclidean distance between the corresponding position and the two state positions is equal, each state position corresponds to a signal state of the balanced signal, and the signal states corresponding to the two state positions are different.

Optionally, the rough judgment module determining the dimensionality reduction position information and the dimensionality reduction signal according to the equalized signal includes: the rough judgment module determines the dimensionality reduction position information according to the equalized signal; the rough judgment module determines the dimensionality reduction signal according to the equalized signal and the dimensionality reduction position information. Wherein, the equalized signal includes the equalized sub-signals at each of the n times, and the rough decision module determines the dimensionality reduction position information according to the equalized signal includes: , to determine the location identification value corresponding to each moment. The coarse judgment module determines the dimensionality reduction signal according to the equalization signal and the dimensionality reduction position information includes: the rough judgment module determines the The dimensionality reduction sub-signal of .

Optionally, when the equalization module includes a first equalization submodule and a second equalization submodule, the coarse decision module determining the dimensionality reduction position information according to the equalization signal includes: the rough decision module determining the dimensionality reduction position information according to the second equalization signal. The coarse judgment module determining the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information includes: the rough judgment module determining the dimensionality reduction signal according to the second equalization signal and the dimensionality reduction location information. That is, the equalized signal used by the rough decision module to determine the dimensionality reduction position information is the second equalized signal, and the equalized signal used by the coarse decision module to determine the dimensionality reduction signal is the first equalized signal. Wherein, when the equalization module includes a first equalization sub-module and a second equalization sub-module, the corresponding position of the dimensionality reduction sub-signal at each moment in the equalization signal is the dimensionality reduction sub-signal at each moment in the second equalization signal corresponding position. For the corresponding position of the dimensionality reduction sub-signal at each of the n times in the second equalized signal, the second equalized signal has two state positions adjacent to the corresponding position, and the corresponding position is located in the two Between the state positions, the corresponding position is equal to the Euclidean distance between the two state positions, each state position in the two state positions corresponds to a signal state of the second balanced signal, and the signal corresponding to the two state positions The status is different.

S1603. The filtering module performs filtering processing on the dimensionality reduction signal to obtain a filtered signal.

The filter module has a filter coefficient, and the filter module can perform filter processing on the dimensionality reduction signal according to the filter coefficient to obtain a filter signal. Wherein, the filtered signal includes filtered sub-signals at each of the n times. In the process of filtering the dimensionality reduction signal, the filtering module converts the colored noise in the dimensionality reduction signal into white noise, and introduces controllable ISI into the dimensionality reduction signal, so the filtered signal contains white noise and controllable ISI.

S1604. The MLSE module determines a target path branch of the filtered signal.

The MLSE module can determine the target path branch of the filtered signal according to the filter coefficient of the filter module. The target path branch includes decision values at each of the n times, and the decision value at each of the n times is obtained by the MLSE module based on the filtered sub-signal at each time.

Wherein, the filtered signal includes the filtered sub-signals at each of the n times, and the MLSE module determines the target path branch of the filtered signal according to the filter coefficient of the filter module, including: the MLSE module determines the target path branch of the filter signal according to the filter coefficient and the n times The filtered sub-signal at each moment in determines the branch metric at each moment; the MLSE module determines the target path branch of the filtered signal according to the branch metrics at n moments.

S1605. The mapping module maps the target path branch of the filtered signal to obtain the target signal according to the dimensionality reduction position information, the number of signal states of the target signal is equal to the number of signal states of the balanced signal, and each signal state of the target signal is equal to the number of signal states of the balanced signal. The respective signal states are the same or correspond.

Wherein, the target path branch of the filtered signal includes a decision value at each of the n moments, and the decision value at each of the n moments is obtained by the MLSE module based on the filtered sub-signal at each moment, The target signal includes target sub-signals at each of the n times. The mapping module maps the target path branch of the filtered signal to obtain the target signal according to the dimensionality reduction position information, including: the mapping module determines the The target sub-signal at each moment; the mapping module determines the target signal according to the target sub-signals at n times.

In the embodiment of the present application, the decision value in the target path branch is a hard value or a soft value. Optionally, when the decision value in the target path branch is a hard value, the mapping module determines the target at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment The sub-signal includes: the mapping module determines the sum of the decision value at each of the n times and the corresponding position identification value at each time as the target sub-signal at each time. When the decision value in the target path branch is a soft value, the mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment, including : The mapping module determines the mapping rule corresponding to each moment according to the position identifier value corresponding to each moment in the n times; the mapping module corresponds to the position identifier value corresponding to each moment according to the judgment value of each moment The mapping rules of , determine the target sub-signal at each moment.

For example, the signal states of the equalized signal include -3, -1, 1, and 3, the signal states of the dimensionality reduction signal include -1 and 1, and the position identifier values include -2, 0, and 2. When the decision value in the target path branch is a soft value, the mapping rule can be as follows:

When p_ind_d(k)=-2, the mapping rule is hd_out(k)=[max, -ld_out(k)];

When p_ind_d(k)=0, the mapping rule is hd_out(k)=[ld_out(k), max];

When p_ind_d(k)=2, the mapping rule is hd_out(k)=[-max, ld_out(k)];

Among them, p_ind_d(k) is the position identification value corresponding to the kth moment, ld_out(k) is the judgment value at the kth moment, hd_out(k) is the target sub-signal at the kth moment, max is a constant, and k is not greater than n A positive integer of , in the expression [A, B] of the mapping rule, A is the high-order bit of the target sub-signal, and B is the low-order bit of the target sub-signal.

Optionally, the signal processing device further includes an autocorrelation estimation module, and the autocorrelation estimation module is respectively connected to the equalization module, the filtering module and the MLSE module. The method also includes: the autocorrelation estimation module determines the filter coefficient of the filter module according to the equalized signal, and provides the filter coefficient to the filter module and the MLSE module respectively, so that the filter module performs filtering processing on the dimensionality reduction signal according to the filter coefficient, and, MLSE The module determines the target path branch of the filtered signal according to the filter coefficient.

Optionally, the signal processing device further includes a decoding module connected to the mapping module. The signal processing method further includes: the decoding module decodes the target signal to recover data.

In summary, in the signal processing method provided by the embodiment of the present application, the equalization module performs equalization processing on the initial signal to obtain an equalized signal, and the rough judgment module determines the dimensionality reduction position information and dimensionality reduction signal according to the equalization signal, and outputs the dimensionality reduction position information To the mapping module, output the dimension reduction signal to the filter module, the filter module filters the dimension reduction signal to obtain the filter signal, and outputs the filter signal to the MLSE module, the MLSE module determines the target path branch of the filter signal, and the target path of the filter signal The path branch is output to the mapping module, and the mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain the target signal. Since the dimensionality reduction signal has nothing to do with the modulation format, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalized signal, and the number of signal states of the filtered signal is equal to the number of signal states of the dimensionality reduction signal, so the filtered signal input to the MLSE module The number of signal states of is less than the number of signal states of the equalized signal output by the equalization module, and the filtered signal has nothing to do with the modulation format, so that the MLSE module needs to calculate fewer path branches in the process of determining the target path branch, and the MLSE module determines the target path The branching process has nothing to do with the modulation format, which helps to reduce the algorithm complexity of the MLSE module and reduce the resource consumption of the MLSE module.

An embodiment of the present application provides a processing chip, and the processing chip includes the signal processing device 10 provided in the foregoing embodiment. Optionally, the processing chip is a DSP chip. Wherein, the processing chip may include a programmable logic circuit and/or program instructions, which are used to implement the signal processing method provided in the above-mentioned embodiments when the processing chip is running.

An embodiment of the present application provides a signal receiving device, where the signal receiving device includes a storage chip and a processing chip. Memory chips are used to store computer programs. The processing chip is configured to execute the computer program stored in the storage chip so that the signal receiving device executes the signal processing method provided by the above embodiments.

Optionally, the signal receiving device is an optical receiver.

An embodiment of the present application provides a signal transmission system, and the signal transmission system includes a signal sending device and the signal receiving device provided in the foregoing embodiments. The signal sending device is connected to the signal receiving device through a transmission link, and the signal sending device is used to send a signal to the signal receiving device through the transmission link.

Optionally, the signal sending device is an optical transmitter, the signal receiving device is an optical receiver, and the transmission link is an optical link. Optical transmitters and optical receivers are collectively referred to as optical communication equipment.

An embodiment of the present application provides a computer-readable storage medium, and a computer program is stored in the computer-readable storage medium. When the computer program is executed (for example, executed by a processing chip), the signal processing method provided in the foregoing embodiments is implemented.

An embodiment of the present application provides a computer program product, where the computer program product includes a program or code, and when the program or code is executed (for example, executed by a processing chip, etc.), the signal processing method provided in the foregoing embodiment is implemented.

In the above embodiments, all or part of them may be implemented by software, hardware, firmware or any combination thereof. For example, the above-mentioned equalization module can be realized through FFE, DFE, etc., the above-mentioned filtering module can be realized through a post-post filter, the above-mentioned decoding module can be realized through FEC, and the above-mentioned coarse judgment module and mapping module can be realized through firmware solidified with program instructions. function etc. When implemented using hardware or firmware, the hardware or firmware includes programmable logic circuits and/or program instructions. When the program instructions are loaded and executed on the optical communication device, the processes or functions according to the embodiments of the present application are generated in whole or in part. The program instructions can be fixed in the processing circuit of the optical communication device. When implemented in software, it may be implemented in whole or in part in the form of a computer program product comprising one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the processes or functions according to the embodiments of the present application will be generated in whole or in part. When the signal processing device provided by the above-mentioned embodiment executes the above-mentioned signal processing method, the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional modules according to needs, that is, The internal structure is divided into different functional modules to complete all or part of the functions described above. In addition, the embodiments of the signal processing device, the processing chip, the signal receiving device, and the signal transmission system provided in the above embodiments belong to the same idea.

Different types of embodiments such as method embodiments and device embodiments provided in the embodiments of the present application may refer to each other. The sequence of operations in the method embodiments provided in the embodiments of this application can be adjusted appropriately, and the operations can also be increased or decreased according to the situation. Any person familiar with the technical field can easily think of changes within the technical scope disclosed in this application. Methods should be covered within the scope of protection of the present application, so they will not be repeated here.

In this application, the term "at least one" means one or more, and "plurality" means two or more, unless otherwise clearly defined. The terms "first" and "second", etc. are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The term "and/or" is only an association relationship describing associated objects, indicating that there may be three types of relationships, for example, A and/or B may indicate: A exists alone, A and B exist simultaneously, and B exists independently. situation. In addition, the character "/" in this article generally indicates that the contextual objects are an "or" relationship.

The above is only an exemplary embodiment of the application, but the protection scope of the application is not limited thereto. Any person familiar with the technical field can easily think of various equivalents within the technical scope disclosed in the application. The modifications or replacements should be covered by the scope of protection of this application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

A signal processing device, characterized in that the signal processing device comprises: an equalization module, a rough decision module, a filter module, a maximum likelihood sequence estimation MLSE module and a mapping module connected in sequence, and the rough decision module is also connected with the map module connection;

The equalization module is used to equalize the initial signal input to the equalization module to obtain an equalized signal;

The rough judgment module is used to determine the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalization signal, and the dimensionality reduction signal includes Dimensionality reduction sub-signals at each of the n moments, the dimensionality reduction signal is independent of the modulation format of the equalized signal, the dimensionality reduction position information includes a position identification value corresponding to each of the n moments, The position identification value corresponding to each moment is used to indicate the corresponding position of the dimensionality reduction sub-signal at each moment in the equalized signal, and n is a positive integer;

The filtering module is used to filter the dimensionality reduction signal to obtain a filtered signal;

The MLSE module is used to determine a target path branch of the filtered signal;

The mapping module is configured to map the target path branch of the filtered signal according to the dimensionality reduction position information to obtain a target signal, the number of signal states of the target signal is equal to the number of signal states of the equalized signal, Each signal state of the target signal is the same as or corresponds to each signal state of the equalized signal.
The signal processing device according to claim 1, wherein:

For the corresponding position of the dimensionality reduction sub-signal at each moment in the equalized signal, the equalized signal has two state positions adjacent to the corresponding position, and the corresponding position is located at the two state positions Between, the Euclidean distance between the corresponding position and the two state positions is equal, each of the state positions corresponds to a signal state of the balanced signal, and the signal states corresponding to the two state positions are different.
The signal processing device according to claim 2, wherein the rough judgment module is specifically used for:

determining the dimensionality reduction position information according to the equalized signal;

The dimensionality reduction signal is determined according to the equalization signal and the dimensionality reduction location information.
The signal processing device according to claim 3, wherein:

The balanced signal includes balanced sub-signals at each of the n times,

The determining the dimension-reduced position information according to the equalized signal includes: determining the position identification value corresponding to each time according to the equalized sub-signal at each time in the n times;

The determining the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information includes: according to the equalization sub-signal at each moment in the n times and the location identification value corresponding to each moment, The dimensionality reduction sub-signal at each moment is determined.
The signal processing device according to claim 3 or 4, characterized in that,

The equalization module includes: a first equalization submodule and a second equalization submodule, the first equalization submodule, the second equalization submodule and the rough decision module are connected in sequence, and the first equalization submodule also connected with the rough judgment module;

The first equalization submodule is used to perform a first equalization process on the initial signal to obtain a first equalized signal;

The second equalization sub-module is configured to perform a second equalization process on the first equalized signal to obtain a second equalized signal, and the equalized signal obtained by the equalized module includes the first equalized signal and the second equalized signal. An equalized signal, the corresponding position of the dimensionality reduction sub-signal in the equalized signal is the corresponding position of the dimensionality reduction sub-signal in the second equalized signal, and the equalized sub-signal is the corresponding position in the second equalized signal sub-signal.
The signal processing device according to claim 5, wherein:

The first equalization submodule is a feed-forward equalizer FFE or a multiple-input multiple-output MIMO equalizer;

The second equalization sub-module is a decision feedback equalizer DFE.
The signal processing device according to claim 5 or 6, wherein the rough decision module is specifically used for:

determining the dimensionality reduction position information according to the second equalized signal;

The dimensionality reduction signal is determined according to the first equalization signal and the dimensionality reduction location information.
The signal processing device according to any one of claims 1 to 7, characterized in that,

The filtered signal includes filtered sub-signals at each of the n moments, the target path branch includes decision values at each of the n moments, and each of the n moments The decision value of is obtained by the MLSE module based on the filtered sub-signals at each moment, and the target signal includes target sub-signals at each of the n moments; the mapping module is specifically used for:

Determine the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment;

The target signal is determined according to the target sub-signals at the n moments.
The signal processing device according to claim 8, wherein:

The judgment value is a hard value, and the mapping module is specifically configured to: determine the sum of the judgment value at each time in the n times and the position identification value corresponding to each time as the sum of each target sub-signal at time.
The signal processing device according to claim 8, wherein:

The decision value is a soft value, and the mapping module is specifically configured to: determine the mapping rule corresponding to each moment according to the position identification value corresponding to each moment in the n times; The decision value of and the mapping rule corresponding to each moment determine the target sub-signal at each moment.
The signal processing device according to claim 10, wherein:

The signal state of the equalized signal includes -3, -1, 1, and 3, the signal state of the dimensionality reduction signal includes -1 and 1, and the position identification value includes -2, 0, and 2;

When p_ind_d(k)=-2, the mapping rule is hd_out(k)=[max, -ld_out(k)];

When p_ind_d(k)=0, the mapping rule is hd_out(k)=[ld_out(k), max];

When p_ind_d(k)=2, the mapping rule is hd_out(k)=[-max, ld_out(k)];

Wherein, the p_ind_d(k) is the position identification value corresponding to the kth moment, the ld_out(k) is the judgment value at the kth moment, and the hd_out(k) is the target sub-signal at the kth moment , max is a constant, k is a positive integer not greater than n, in the expression [A, B] of the mapping rule, A is the upper bit of the target sub-signal, and B is the lower bit of the target sub-signal.
The signal processing device according to any one of claims 1 to 11, characterized in that,

The signal processing device further includes: an autocorrelation estimation module, the autocorrelation estimation module is respectively connected to the equalization module, the filtering module and the MLSE module;

The autocorrelation estimation module is used to determine the filter coefficient of the filter module according to the equalized signal;

The filtering module is specifically configured to perform filtering processing on the dimensionality reduction signal according to the filtering coefficient;

The MLSE module is specifically configured to determine a target path branch of the filtered signal according to the filter coefficient.
The signal processing device according to claim 12, wherein:

The filtered signal includes filtered sub-signals at each of the n times,

The MLSE module is specifically designed to:

determining the branch metric at each moment according to the filter coefficient and the filter sub-signal at each moment in the n moments;

A target path branch of the filtered signal is determined according to the branch metrics at the n time moments.
The signal processing device according to any one of claims 1 to 13, characterized in that,

The signal processing device further includes: a decoding module connected to the mapping module;

The decoding module is used to decode the target signal to restore data.
The signal processing device according to any one of claims 1 to 14, characterized in that,

The signal processing means is an intensity modulation means or a phase modulation means.
The signal processing device according to any one of claims 1 to 15, characterized in that,

The equalized signal is a complex signal or a real signal.
The signal processing device according to claim 16, wherein:

The equalization signal is a pulse amplitude modulation PAM signal, and the dimensionality reduction signal is a binary amplitude keying OOK signal; or,

The equalized signal is a quadrature amplitude modulation QAM signal, and the dimensionality reduction signal is a quadrature phase shift keying QPSK signal.
A signal processing method, characterized in that it is applied to a signal processing device, and the signal processing device includes an equalization module, a rough decision module, a filter module, a maximum likelihood sequence estimation MLSE module and a mapping module connected in sequence, and the rough decision The module is also connected with the mapping module, and the method includes:

The equalization module performs equalization processing on the initial signal input to the equalization module to obtain an equalized signal;

The rough judgment module determines dimensionality reduction position information and dimensionality reduction signals according to the equalization signal, the number of signal states of the dimensionality reduction signal is less than the number of signal states of the equalization signal, and the dimensionality reduction signal includes n Dimensionality reduction sub-signals at each moment in time, the dimensionality reduction signal has nothing to do with the modulation format of the equalized signal, the dimensionality reduction position information includes the position identification value corresponding to each moment in the n times, each The position identification value corresponding to the time is used to indicate the corresponding position of the dimensionality reduction sub-signal at each time in the equalized signal, and n is a positive integer;

The filtering module performs filtering processing on the dimensionality reduction signal to obtain a filtered signal;

the MLSE module determines a target path branch of the filtered signal;

The mapping module maps the target path branch of the filtered signal according to the dimensionality reduction position information to obtain a target signal, the number of signal states of the target signal is equal to the number of signal states of the equalized signal, and the Each signal state of the target signal is the same as or corresponds to each signal state of the equalized signal.
The method of claim 18, wherein,

For the corresponding position of the dimensionality reduction sub-signal at each moment in the equalized signal, the equalized signal has two state positions adjacent to the corresponding position, and the corresponding position is located at the two state positions Between, the Euclidean distance between the corresponding position and the two state positions is equal, each of the state positions corresponds to a signal state of the balanced signal, and the signal states corresponding to the two state positions are different.
The method of claim 19, wherein,

The rough judgment module determines the dimensionality reduction position information and the dimensionality reduction signal according to the equalization signal, including:

The rough decision module determines the dimensionality reduction position information according to the equalized signal;

The rough decision module determines the dimensionality reduction signal according to the equalization signal and the dimensionality reduction location information.
The method of claim 20, wherein

The balanced signal includes balanced sub-signals at each of the n times,

The rough judgment module determines the dimensionality reduction position information according to the equalized signal, including: the rough judgment module determines the position corresponding to each time according to the equalized sub-signal at each time in the n times identity value;

The rough judgment module determines the dimensionality reduction signal according to the equalization signal and the dimensionality reduction position information, including: the rough judgment module according to the equalization sub-signal at each moment in the n times and the each The position identification value corresponding to each moment, and determine the dimensionality reduction sub-signal at each moment.
The method according to claim 20 or 21, characterized in that,

The equalization module includes a first equalization submodule and a second equalization submodule, the first equalization submodule, the second equalization submodule and the rough decision module are connected in sequence, and the first equalization submodule is also connected with The coarse decision module is connected;

The equalization module equalizes the initial signal input to the equalization module to obtain an equalized signal, including:

The first equalization sub-module performs a first equalization process on the initial signal to obtain a first equalized signal;

The second equalization sub-module performs a second equalization process on the first equalized signal to obtain a second equalized signal, and the equalized signal obtained by the equalized module includes the first equalized signal and the second equalized signal , the corresponding position of the dimensionality reduction sub-signal in the equalized signal is the corresponding position of the dimensionality reduction sub-signal in the second equalized signal, and the equalized sub-signal is a sub-signal in the second equalized signal Signal.
The method of claim 22, wherein,

The rough judgment module determining the dimensionality reduction location information according to the equalized signal includes: the rough judgment module determining the dimensionality reduction location information according to the second equalization signal;

The rough judgment module determines the dimensionality reduction signal according to the equalization signal and the dimensionality reduction position information, including: the rough judgment module determines the dimensionality reduction signal according to the first equalization signal and the dimensionality reduction position information Signal.
A method according to any one of claims 18 to 23, wherein,

The filtered signal includes filtered sub-signals at each of the n moments, the target path branch includes decision values at each of the n moments, and each of the n moments The decision value of is obtained by the MLSE module based on the filtered sub-signal at each moment, and the target signal includes the target sub-signal at each moment in the n moments;

The mapping module maps the target path branch of the filtered signal to obtain the target signal according to the dimensionality reduction position information, including:

The mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment;

The mapping module determines the target signal according to the target sub-signals at the n time points.
The method of claim 24, wherein,

The decision value is a hard value, and the mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment, including :

The mapping module determines the sum of the decision value at each of the n moments and the position identification value corresponding to each moment as the target sub-signal at each moment.
The method of claim 24, wherein,

The decision value is a soft value, and the mapping module determines the target sub-signal at each moment according to the decision value at each moment in the n moments and the position identification value corresponding to each moment, including :

The mapping module determines the mapping rule corresponding to each moment according to the position identification value corresponding to each moment in the n times;

The mapping module determines the target sub-signal at each time according to the decision value at each time and the mapping rule corresponding to the location identification value corresponding to each time.
The method of claim 26, wherein

The signal state of the equalized signal includes -3, -1, 1, and 3, the signal state of the dimensionality reduction signal includes -1 and 1, and the position identification value includes -2, 0, and 2;

When p_ind_d(k)=-2, the mapping rule is hd_out(k)=[max, -ld_out(k)];

When p_ind_d(k)=0, the mapping rule is hd_out(k)=[ld_out(k), max];

When p_ind_d(k)=2, the mapping rule is hd_out(k)=[-max, ld_out(k)];

Wherein, the p_ind_d(k) is the position identification value corresponding to the kth moment, the ld_out(k) is the judgment value at the kth moment, and the hd_out(k) is the target sub-signal at the kth moment , max is a constant, k is a positive integer not greater than n, in the expression [A, B] of the mapping rule, A is the upper bit of the target sub-signal, and B is the lower bit of the target sub-signal.
A method according to any one of claims 19 to 27, wherein

The signal processing device also includes an autocorrelation estimation module, the autocorrelation estimation module is respectively connected to the equalization module, the filtering module and the MLSE module;

The method further includes: the autocorrelation estimation module determines the filter coefficient of the filter module according to the equalized signal;

The filtering module performs filtering processing on the dimensionality reduction signal, including: the filtering module performs filtering processing on the dimensionality reduction signal according to the filtering coefficient;

The MLSE module determining the target path branch of the filtered signal includes: the MLSE module determining the target path branch of the filtered signal according to the filter coefficient.
The method of claim 28, wherein,

The filtered signal includes filtered sub-signals at each of the n times,

The MLSE module determines the target path branch of the filtered signal according to the filter coefficient, including:

The MLSE module determines the branch metric at each moment according to the filter coefficient and the filter sub-signal at each moment in the n moments;

The MLSE module determines a target path branch of the filtered signal according to the branch metrics at the n moments.
A method according to any one of claims 18 to 29, wherein,

The signal processing device further includes a decoding module connected to the mapping module;

The method further includes: the decoding module decodes the target signal to recover data.
A processing chip, characterized by comprising the signal processing device according to any one of claims 1 to 17.
The processing chip according to claim 31, characterized in that, the processing chip is a digital signal processing DSP chip.
A signal receiving device, characterized in that it includes a memory chip and a processing chip;

The memory chip is used to store computer programs;

The processing chip is used to execute the computer program stored in the storage chip so that the signal receiving device executes the signal processing method according to any one of claims 18 to 30.
A signal transmission system, characterized in that it comprises: a signal sending device and a signal receiving device according to claim 33, the signal sending device is connected to the signal receiving device through a transmission link, and the signal sending device uses for sending a signal to the signal receiving device through the transmission link.
The signal transmission system according to claim 34, wherein the signal sending device is an optical transmitter, the signal receiving device is an optical receiver, and the transmission link is an optical link.
A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed, the signal processing method according to any one of claims 18 to 30 is implemented.
A computer program product, characterized in that the computer program product includes a program or code, and when the program or code is executed, the signal processing method according to any one of claims 18 to 30 is realized.