WO2015103804A1

WO2015103804A1 - Channel equalization and frequency offset estimation joint parallel method based on lms

Info

Publication number: WO2015103804A1
Application number: PCT/CN2014/072120
Authority: WO
Inventors: 吴晨雨; 许渤; 刘芯羽
Original assignee: 电子科技大学
Priority date: 2014-01-07
Filing date: 2014-02-16
Publication date: 2015-07-16
Also published as: CN103780519A; CN103780519B

Abstract

Disclosed is a channel equalization and frequency offset estimation joint parallel method based on an LMS. After a training sequence signal is converted into a parallel signal in an initialization stage of a photoreceiver, the signal is sent to a parallel signal processing branch, and equalization and frequency offset estimation are respectively performed on the signal; a frequency offset estimation average value of all the branches and an error signal of each branch are calculated; each group of parallel signals use a unified equalizer tap coefficient, and an average value of the branch error signals is used when the equalizer tap coefficient is updated; a sending end inserts a training symbol into a data symbol in a data sending stage; the photoreceiver converts a data signal into a parallel signal; a training signal uses an equalization signal and the known training symbol to perform frequency offset estimation; and the data signal uses an accumulated phase error of a corresponding training signal and the obtained frequency offset estimation average value to compensate and then to judge, and then uses the equalization signal and a judgement signal to perform frequency offset estimation. By employing parallel signal processing, the present invention reduces the limitations and influence of data signal processing hardware on the system performance.

Description

LMS-based joint equalization method for channel equalization and frequency offset estimation based on LMS

The invention belongs to the technical field of optical burst receivers, and more particularly to a joint parallelization method based on LMS (Least Mean Square) for channel equalization and frequency offset estimation. Background technique

In the current high-speed coherent optical communication system, PDM-QPSK (polarization multiplexing-four-phase absolute phase shift keying) coherent optical transmission system is one of the most promising technical solutions. In the PDM-QPSK coherent optical transmission system, the transmission signal is mainly affected by the linear damage of the Chromatic Dispersion (CD) and the Polarization Mode Dispersion (PMD) of the fiber and the frequency offset generated by the transceiver laser. Impact, these two problems seriously affect the performance of the optical receiver. The adaptive equalization technique can basically eliminate crosstalk between codes caused by dispersion, and the frequency offset estimation technique can be used to solve the effect of frequency offset. Since the equalizer and the frequency offset estimator interact with each other, a joint algorithm of time domain equalization and frequency offset estimation can be used.

For optical burst transmission systems, the characteristics of optical bursts require that the equalizer in the optical burst receiver be capable of fast convergence. Figure 1 is a system block diagram of an optical burst receiver. As shown in Fig. 1, the input signal 0 of the receiver is a signal in which two PDM-QPSK signals whose polarization directions are perpendicular to each other are polarization-coupled and transmitted over a certain distance of the optical fiber channel. During the transmission of the Fibre Channel, the optical signal is affected by factors such as dispersion, polarization mode dispersion, and optical amplifier noise, resulting in degradation of the quality of the transmitted signal. The PDM-QPSK signal r(t) enters the 90-degree mixer with the FTLO (fast tunable laser) light wave for coherent demodulation. The four signals after coherent demodulation are subjected to AD sampling and quantization. The sampled and quantized 4-channel signals /x, Qx, ly, respectively represent the in-phase and quadrature-modulated signals of the two polarization states x, _y, which enter the digital signal processing module for channel equalization. And Frequency Offset Estimation (FOE) and compensation, and the final phase decision recovers the transmitted data.

A joint algorithm based on LMS for channel equalization and frequency offset estimation is an effective method to improve the performance of coherent optical receivers. However, in the design of an optical burst receiver, when a digital signal processing algorithm is implemented using an FPGA (Field-Programmable Gate Array) or an application-specific integrated circuit, The calculation speed and chip area are two main problems that are mutually constrained. Therefore, it is necessary to choose between performance and implementation complexity. Due to the high rate characteristics of fiber-optic communication, the 112Gb/s PDM-QPSK fiber transmission system is taken as an example. The symbol rate of each of the four signals after coherent demodulation is 28G/S, and the four branch electrical signals need to be first. The rate of AD sampling and quantization, each branch signal rate is up to 56G / S, so the symbol entering the equalizer is a discrete signal with a symbol rate of 56G / S. The subsequent digital signal processing unit (DSPU) cannot handle the rate in hardware, so parallel processing must be used. Depending on the rate of input data and the processing speed of the chip, it is possible to use larger values for the number of parallel branches. This requires that in real-time applications, the algorithm must meet the requirements of parallel processing. At the same time, the update of the equalizer tap coefficients and the frequency offset estimation algorithm based on the pre-decision all require signal feedback. In the parallel implementation, the delay caused by the feedback has a great influence on the performance of the system. Therefore, in the specific implementation design of the optical burst receiver, the effects of parallel and feedback delay on receiver performance must also be considered. Summary of the invention

The object of the present invention is to overcome the deficiencies of the prior art and provide a joint parallel method for channel equalization and frequency offset estimation based on LMS, which reduces the influence of data signal processing hardware on system performance.

To achieve the above object, the present invention is based on an LMS-based joint equalization method for channel equalization and frequency offset estimation, including the following steps:

S1: Initialization using a training sequence, including steps:

S1.1: transmitting the training sequence to the optical burst receiver, performing the coherent demodulation and sampling and quantizing the training sequence signal to perform serial-to-parallel conversion to obtain the N-channel parallel signal; setting the equalizer tap coefficient corresponding to the M=l group parallel signal

S1.2: The first group of parallel signals enters N parallel signal processing branches, each parallel signal processing branch includes an equalizer and a frequency offset estimation module, and the equalizer of the first branch obtains an equalized signal ^ = 3⁄4, where k = (n - l) x N + i , \ ≤ i ≤ N ;

S1.3: The frequency offset estimation module performs frequency offset estimation on the equalized signal 3⁄4 according to the known training symbol _e , and obtains a cumulative phase error “3⁄4 and a frequency offset estimation value Δ „ ;

S1.4: averaging the frequency offset estimates Δ „ of the N branches to obtain the average of the frequency offset estimates of the "group" parallel signals Δ Γ ;

S1.5: N branches calculate their error signals separately

ε _η . = m _k ― e ^j9k · e ^JA& · e ^j(p " S1.6: Update the equalizer tap coefficients used for the M + l group of parallel signals:

Where „ ₊₁ , respectively represent the equalizer tap coefficients used in the “+ 1 group, the first group” parallel signal; the delay indicating the error signal; the set iteration length, which is a positive number; represents the N branches The number of error signals selected to participate in the tap coefficient calculation, 1 ≤ ≤ N, 1 < _C ≤ N _C ; V(n - D, i _c ) denotes the observation vector corresponding to _DJc , f(n - denotes f(n - Conjugation of D, i _c );

S1.7: determining whether the training sequence is processed, if not processed, returning to step S 1.2 to continue processing the next set of parallel signals, if the processing is completed, proceed to step S2;

S2: Enter the data transmission stage to process the data, including steps:

S2.1: The data transmitting end inserts a training symbol into the data symbol, and the insertion method is: grouping N sending symbols into groups, and then dividing the N sending symbols into R groups, and each group of sending symbols includes one training. Symbol, the serial number of the R training symbols in the parallel symbol is recorded as i ≤ r ≤ w;

S2.2: transmitting a data signal to the optical burst receiver, performing serial-to-parallel conversion on the coherent demodulation and sampling and quantizing data signals to obtain N parallel signals, and the “group parallel signals enter N parallel signal processing branches, data The parallel signal processing branch of the transmitting phase includes an equalizer, a frequency offset estimation module and a decision module, and the equalizer of the first branch processes the equalized signal

S2.3: Perform frequency offset estimation on each of the N branches:

When the branch is a training signal, the frequency offset estimation is performed on the equalized signal ^ directly according to the known training symbol _e , and the cumulative phase error "3⁄4 and the frequency offset estimated value Δ „ is obtained; when the branch is a data signal, the equalized signal is first used. 3⁄4 for phase compensation, the phase compensated signal 3⁄4 is:

Where ^ represents the delay of the average value of the frequency offset estimation, "·1 indicates rounding up; the decision module decides the signal to obtain the decision signal, and performs frequency offset estimation on the equalized signal 3⁄4 according to the decision signal to obtain the frequency offset estimated value Δ „ and Cumulative phase error <3⁄4;

S2.4: Average the frequency offset estimates Δώ„ of one branch to obtain the frequency offset of the nth parallel signal Estimated mean Δ Γ;

S2.5: N branches calculate their error signals separately

When the branch is a training signal, g^' =, the error signal is:

ε _η . =m _k - e ^J9k · e ^JA& · e ^j( "

When the branch is a data signal, the error signal ^ is:

ε _η . = m _k -e ^A · e ^J Ά ¹

S2.6: Update the η + 1 group tap coefficients:

Where Γ is the iteration length set in the data transmission phase, N: represents the number of error signals calculated by the participating tap coefficients selected from the N branches in the data transmission phase, i ≤ N: ≤ N,

S2.7: Determine whether the data has been processed. If it has not been processed, return to step S2.2 to continue processing. If the processing is completed, the process ends.

Further, the specific method of frequency offset estimation includes the following steps:

S3.1: Calculate the cumulative phase error of the equalized signal 3⁄4:

When the branch is a training symbol, the cumulative phase error "3⁄4 = Φ Α _ , where the phase of the known training symbol is represented, represents the phase of the equalized signal 3⁄4;

When the branch is a data signal, the accumulated phase error is <3⁄4 = 3^_, where 4 represents the phase of the data decision signal e ⁱ ;

S3.2: Calculate the frequency offset estimate of this branch ΔώΤ^^ - ^ where represents the cumulative phase error of the A-1th signal.

The invention is based on the LMS-based channel equalization and frequency offset estimation joint parallel method. In the initialization stage of the optical receiver, the sampling signal of the training sequence signal is converted into a parallel signal by serial-to-parallel conversion and then sent to the parallel signal processing branch, each branch The equalization and frequency offset estimation are performed respectively, and the frequency offset estimates obtained by all the branches are averaged to obtain the average of the frequency offset estimation. Each branch calculates the error signal according to the average value of the frequency offset estimation, and each group of parallel signals is unified. The equalizer tap coefficient is updated by the mean value of the branch error signal when the equalizer tap coefficient is updated; in the data transmission phase, the transmitting end inserts the training symbol into the data symbol, and passes the coherent demodulation and the sampled quantized data signal through the string. And transform to obtain a parallel signal, the training signal uses the equalized signal and the known training symbols for frequency offset estimation, and the data signal uses corresponding The accumulated phase error of the training signal and the obtained average of the frequency offset estimation are compensated and then judged, and then the equalization signal and the decision signal are used for frequency offset estimation, and then the equalizer tap coefficient is updated in the same manner as the initialization phase.

The invention has the following beneficial effects:

(1) The present invention reduces the signal rate by parallelization, thereby reducing the influence of data signal processing hardware on system performance;

(2) In the initialization phase, the error signal is calculated by using the average value of the frequency offset estimation, which can improve the reliability of the equalizer initialization;

(3) In the data transmission phase, inserting a certain number of training symbols into the parallel symbols can improve the accuracy of the decision, frequency offset estimation and error signal feedback;

(4) The simulation shows that the invention has good tolerance to the delay of the error signal. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a system block diagram of an optical burst receiver;

2 is a schematic diagram of a parallel signal branch algorithm in an initialization phase;

3 is a schematic diagram of a parallel signal branch algorithm in a data transmission phase;

4 is a comparison diagram of equalizer convergence speeds of the parallel method and the serial method of the present invention;

Figure 5 is a comparison of the convergence speeds of the computational delay and the no computational delay in the parallel method of the present invention; Figure 6 is a comparison of the bit error rate of the serial method and the parallel method of the different delays of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The detailed description of the embodiments of the present invention will be described in the claims It is to be noted that in the following description, when a detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted herein. Example

In this embodiment, the 112 Gb/s PDM-QPSK optical fiber transmission system is taken as an example, and the 56 G/s rate signal is serially converted into multiple parallel signals, here 256 channels, so that the rate of each channel can be effectively reduced. , which can reduce the difficulty of its physical implementation. Since it is doubled, each group of 256 signals will result in 128 symbols, and the number of parallel processing branches that need to be configured is N = 128. The operation of an optical burst receiver is divided into two phases: an initialization phase and a data transmission phase. After detecting the arrival of the optical burst signal, the optical burst receiver first enters the initialization phase, and the training sequence is used to iteratively update the equalizer tap coefficients until convergence, and the initialization of the equalizer and the optical burst receiver is completed. After the initialization of the optical burst receiver is completed, the data transmission phase is entered. The algorithm of the two stages of the present invention will be described in detail below.

First, the initialization phase

Figure 2 is a schematic diagram of the parallel signal branch algorithm in the initialization phase. As shown in FIG. 2, the parallel algorithm of the present invention is different from the serial algorithm in that a set of 128 parallel symbols corresponds to 128 LMS algorithm-based equalizers (EQs), each equalizer adopts the same The tap coefficient. The update of the tap coefficients is the key to the initialization of the equalizer. The update of the tap coefficients requires the use of an error signal, so the error signal for each branch needs to be obtained first. The training sequence used in the initialization phase should be long enough to ensure that the equalizer tap coefficients obtained by the initialization can converge. The initialization phase includes the following specific steps:

S101: Send a training sequence to the optical burst receiver, and perform serial-to-parallel conversion on the training sequence signal of the coherent demodulation and the sample quantization to obtain an N-channel parallel signal; and set an equalizer tap coefficient corresponding to the M=l group parallel signal. In this embodiment, it is 128 channels.

S102: The first group of parallel signals enters N parallel signal processing branches, each parallel signal processing branch includes an equalizer and a frequency offset estimation module, and the equalizer of the first branch obtains an equalized signal ^=3⁄4, where k=( n - l) x N + i , \ ≤ i ≤ N.

S103: The frequency offset estimation module performs frequency offset estimation on the equalization signal 3⁄4 according to the known training symbol _e , to obtain a cumulative phase error 3⁄4 and a frequency offset estimation value Δ „.

The frequency offset estimation method used in this embodiment is a phase estimation method based on pre-decision, and the algorithm considers: The phase of the equalized signal ^ can be expressed as: where is the phase information carried by the symbol, and "3⁄4 is the cumulative phase error. The phase error "3⁄4 can be expressed as:

Among them, it is caused by laser phase noise, which is slowly changing for high-speed signals, so it can be considered as a constant for a group of parallel symbols, and is the phase fluctuation caused by ASE (spontaneous radiation) noise. Caused by partiality. It can be seen that after the phase of each signal is removed, the phase of the symbol is removed. The remaining "3⁄4, differential operation of the cumulative phase error of adjacent symbols, to obtain the frequency offset estimate of each branch

+ Α, and then perform the averaging operation to obtain the average value of the frequency offset estimation Δώ„Γ, which can suppress the influence to some extent. The frequency offset is slowly changing for the high-rate signal, so in the present invention, for a group of parallel The symbol can be considered to have the same frequency offset. The specific steps of the frequency offset estimation include:

S3.1: Calculate the cumulative phase error of the equalized signal 3⁄4 : ^^-^, where ^ ^ denotes the phase of the equalized signal 3⁄4.

S3.2: Calculate the frequency offset estimate of the branch Δώ^Τ^ - ^ where ^ denotes the cumulative phase error of the first -1 symbol of the training sequence. Obviously, when the first branch in the initialization phase calculates the frequency offset estimate, the initial value of the accumulated phase error is “3⁄4 =0.

In this embodiment, as shown in FIG. 2, step S3.1 and step S3.2 are obtained by multiplying the equalization signal ^ with the conjugate ("^ (·)) of the training sequence symbol e to obtain ₆ ^, _The multiplication of _e ^ with the conjugate gives ^, and the angle of the angle (arg( ) is obtained to obtain the frequency offset estimate Δώ„.

S104: averaging the frequency offset estimation results of the N branches by Δώ„ to obtain the frequency of the “group parallel signal”

1 Ν

The partial estimate is §ρ _Δ( 3⁄4Γ = ∑Δώ„ Γ.

S105: N branches calculate their error signals separately

ε _η . =m _k - e ^J9k · e ^JA& · e ^j( "

Since the known training sequence is used in the initialization phase, it is not necessary to use the decision signal when calculating the error signal, but the known training symbols are directly used.

S106: Update the equalizer tap coefficients used by the M+1 group of parallel symbols:

Where d „ ₊₁ represents the equalizer tap coefficients used in the “+ i group and the “group” parallel signals respectively. It represents the delay of the error signal, that is, the time when the parallel symbol is input to the error signal and fed back to the equalizer. Delay, so when 1 ≤ M ≤, the tap coefficients cannot be updated, and the tap coefficients always use the initial value ^. It is the set iteration length, which is a positive number, and its selection needs to be small enough to ensure that the iterative process can converge. . The equalizer used in the present invention is an equalizer based on the LMS algorithm. When updating the tap coefficients, it is not a single-path error signal, but the mean value of the branch error signal, that is, -V(n - D,i ] , N. indicates the error of the participation in the calculation of the tap coefficients selected from the N branches

The number of signals, 1 ≤ ≤ W, i _c ≤ N _c . When the number of branches is large, all the calculation of a set of error signals will generate a large delay, so under the condition that convergence can be satisfied, the number of error signals participating in the average calculation can be reduced, that is, < W. f(M -A represents the corresponding observation vector, that is, the signal input to the equalizer, the conjugate of the representation.

S107: Determine whether the training sequence signal is processed or not. If the processing is not completed, the processing returns to step S102 to continue processing the next set of parallel signals. If the processing is completed, the data transmission phase is entered.

Second, the data transmission phase

The main difference between the data transmission phase and the initialization phase is that the phase of the data symbol is unknown and needs to be recovered by decision. At the same time, since the phase decision process of the data symbols may be erroneous, the present invention inserts a certain number of training symbols into the transmitted data symbols to obtain the reference phase required for phase compensation. Figure 3 is a schematic diagram of the parallel signal branch algorithm in the data transmission phase. As shown in Figure 3, the data transmission phase includes the following steps:

S201: The data sending end inserts the training symbol into the data symbol, and the insertion method is: grouping the N sending symbols into groups, and then dividing the N sending symbols into R groups, and each group of sending symbols includes a training symbol. The serial numbers of the R training symbols in the parallel symbols are denoted by i ≤ ^ ≤ w.

In the data transmission phase, since each group of N symbols after parallel needs to pre-judicate N branches at the same time, if the error occurs in the judgment, the error signal of the feedback is likely to be wrong, which will have a bad influence on the performance of the system. Therefore, in order to obtain a relatively accurate phase compensation when making a decision, the present invention inserts a certain number of training symbols into the transmitted symbols when the transmitting end transmits a signal. The branch corresponding to the training symbol is the same as the calculation method of each branch in the training sequence stage, and the training signal branch first calculates the accurate cumulative phase as the reference phase used in the determination of the data signal branches before and after the branch, so as to perform More accurate frequency offset estimation.

In order to minimize the noise that is amplified when the frequency offset estimate is made at the time of the decision, a preferred way is to insert the insertion symbol in the middle of the group, taking a group of 128 transmission symbols as an example, then inserting The position of the training symbol is: 128/(Rx2) + xxl28/R, x = 0,1,...,R. For example, when each group inserts 4 symbols, then 32 symbols per group are judged based on the same cumulative phase error, so 4 training symbols are inserted at the same =16, _3⁄4 = 48, ₃ = 80, =11 ² On the branch road, the phase compensation error can be reduced at the time of the decision, and up to 16 times the estimated value of the training signal branch frequency offset is used.

S202: transmitting a data signal to the optical burst receiver, performing serial-to-parallel conversion on the coherent demodulation and sampling and quantizing data signals to obtain N parallel signals, and “the group parallel signal enters N parallel signal processing branches, and the data transmission phase The parallel signal processing branch includes an equalizer, a frequency offset estimation module, and a decision module, and the equalizer of the first branch processes the equalized signal 3⁄4.

The parallel symbol group number "in the data transmission phase" is continuously arranged from the sequence number of the last parallel symbol group in the initialization phase. The equalizer tap coefficient of the first group of parallel symbols in the data transmission phase is the equalization obtained at the end of the initialization phase. Tap coefficient.

S203: Perform frequency offset estimation on each of the N branches. The processing flow of the training signal and the data signal are different.

When the branch is a training signal, the ₁₆ signals shown in FIG. 3 are obtained by the same algorithm as the initialization phase, and SP is used to directly estimate the frequency offset of the equalized signal 3⁄4 according to the known training symbol ^'. The partial estimate and the cumulative phase error <3⁄4, that is, and φ _η ' ₁₆ .

When the branch is a data signal, as shown in Figure 3, the ₁₅ and ₁₇ signals are phase compensated for the equalized signal 3⁄4, and the phase compensated signal 3⁄4 is:

X _d T.(ji _r , -j(p„, _ir

Mk = ^m k

N/R where d is the delay of the average value of the frequency offset estimation, “·1 indicates the rounding up. It can be seen that the phase compensation of the data signal in the “parallel signal of the group” in the data transmission phase uses the frequency offset obtained by the first parallel signal. Estimate the average and the cumulative phase error obtained from the training signals in the group to which it belongs.

15

Taking , 15 as an example, since ^ 1 16, the phase compensated signal ^ 15

N/R 32

X— .(15- 16)— , ₁₆

m _nl5 = m _nl5 .e

]Αώ _η _ _ά Τ-]φ _η

m _{n 5} 'e

The decision module judges the signal to obtain a decision signal, and the equalization is performed according to the decision signal No. 3⁄4 performs frequency offset estimation, and obtains the frequency offset estimation value Δ „ and the cumulative phase error “3⁄4, as shown in Figure 3

S204: averaging the frequency offset estimation results Δώ„ of the N branches to obtain a frequency offset estimation average value Δ τ of the “group parallel signal”, that is,

S205: N branches calculate their error signals respectively. Similarly, the processing methods of training signals and data signals are different.

When the branch is a training signal, g^' =, the error signal is:

ε _η . = m _k - e ^J9k · e ^JA& · e ^j( "

When the branch is a data signal, the error signal ^ is:

£ _ni = _k - e ^{A Δώ} » ^Γ - e ^iVk - ¹

S206: Update the n + 1 group coefficient:

Where Γ is the iteration length set in the data transmission phase, and N: represents the number of error signals calculated by the participating tap coefficients selected from the N branches in the data transmission phase, i ≤ N: ≤ N, \ ≤ i: ≤ . If the number of error signals involved in the average calculation in the initialization phase and the data transmission phase is different from N:, the iteration lengths λ and Γ they use also need to be adjusted accordingly.

In the data transmission phase, since the operating time of the system has exceeded the error signal delay of £), the update of the equalizer tap coefficients can be achieved each time.

S207: Determine whether the data is processed. If the processing is not completed, return to step S202 to continue processing the next set of parallel signals, and if the processing is completed, the process ends.

The following is a simulation verification of the LMS-based channel equalization and frequency offset estimation joint parallel method of the present invention. The number of parallel branches in the simulation is 256. Using standard single-mode fiber, the fiber transmission distance is about 50km, and the equalizer uses 11 taps.

Firstly, the influence of the number of error signals in the update stage tap coefficient update on the system performance is simulated. The calculation delay of the equalizer error signal takes 10 clock units, and the delay of FOE calculation also takes 10 clock units, for a total of 20 delay units. In the initialization phase, 12 frames of training data are used, 1024 symbols per frame, with a duration of approximately 440 ns. In the data transmission phase, each Four training symbols are inserted into the group parallel data. Other parameters used in the simulation include a 1G frequency offset and an optical signal-to-noise ratio (OSNR) fixed at 13 dB. Table 1 shows the effect of the number of different error signals on the system performance in the equalizer tap coefficient update.

Table 1

It can be seen from Table 1 that when the optical signal-to-noise ratio is 13 dB, only 64 error signals are needed to obtain good system performance.

Then, the impact of the number of training symbols inserted in each group of signal insertions in the data transmission phase on the system performance is simulated. Two simulations are performed with OSNR of 12dB and 13dB respectively. The other parameters are the same as those used in Table 1. Table 2 shows the effect of the number of training symbols inserted in each group of transmitted symbols during the data transmission phase on system performance.

Table 2

As can be seen from Table 2, in the case where the OSNR is 12 dB and 13 dB, respectively, the number of inserted training symbols is 4, and sufficient performance can be obtained. When designing a burst optical receiver, the number of different insertion training symbols can be selected according to different system design requirements.

4 is a comparison diagram of equalizer convergence speeds of the parallel method and the serial method of the present invention. The OSNR used in this simulation is 13dB, and MSE stands for Mean Squared Error. As shown in Figure 4, the parallelization of the serial algorithm will reduce the convergence speed of the equalizer to a certain extent, but it will not affect the performance after convergence.

Figure 5 is a graph comparing the convergence speeds of the calculated delay and the no computational delay in the joint parallel method of the present invention. The total delay used in this simulation is 20 clock units. As shown in Figure 5, when there is a computational delay, the convergence of the equalizer is only delayed with the delay of the iterative calculation, and does not affect the performance after the iteration is converged. And if you use other different delay sizes in your simulation, you can get similar effects. Thus, it can be seen that the present invention has a good tolerance for calculating the magnitude of the delay.

6 is a comparison of bit error rates of the serial method and the parallel method of different delays of the present invention. As shown in Fig. 6, the calculated delay size in the present invention has no effect on the bit error rate (BER) performance of the optical burst receiver as long as the optical burst receiver is properly initialized. At the same time, compared with the performance of the ideal serial method, the performance loss due to parallelization in the present invention is only about 0.2 dB.

While the invention has been described with respect to the preferred embodiments of the present invention, it should be understood that These variations are obvious as long as the various changes are within the spirit and scope of the invention as defined and claimed in the appended claims, and all inventions that utilize the inventive concept are protected.

Claims

claims

1. A joint parallel method of channel equalization and frequency offset estimation based on LMS, which is characterized by including the following steps:

S1: Initialization using training sequence, including steps:

S1.1: Send the training sequence to the optical burst receiver. After coherent demodulation and sampling and quantization, the training sequence signal undergoes serial-to-parallel conversion to obtain N parallel signals; set the equalizer tap coefficient corresponding to the M = lth group of parallel signals.

S1.2: The th group of parallel signals enters N parallel signal processing branches. Each parallel signal processing branch includes an equalizer and a frequency offset estimation module. The equalizer of the th branch obtains an equalized signal ^ =¾, where k = (n-l)xN + i , 1 < < N;

S1.3: The frequency offset estimation module estimates the frequency offset of the equalized signal ¾ based on the known training symbol _e , and obtains the cumulative phase error ¾ and the frequency offset estimate Δ „;

S1.4: Average the estimated frequency offset values Δ „ of the N branches to obtain the estimated average frequency offset Δ Γ of the w-th group of parallel signals;

S1.5: N branches calculate their error signals respectively

ε _η . =m _k - e ^J9k · e ^JA& · e ^j( "

S1.6: Update the equalizer tap coefficients used by the M + l group of parallel signals:

Among them, d„ ₊₁ and represent the equalizer tap coefficients used by the "+i" and "+" groups of parallel signals respectively; represents the delay of the error signal; is the set iteration step length, which is a positive number; represents the starting point from N branches The number of error signals selected in the path to participate in the calculation of tap coefficients, 1≤ ≤N, 1< _C ≤N _C ; V(nD, _ic ) represents the observation vector corresponding to _DJc , f(n - represents f(n - D ,i _c ) conjugation;

S1.7: Determine whether the training sequence has been processed. If not, return to step S1.2 to continue processing the next set of parallel signals. If the processing is completed, enter step S2;

S2: Enter the data sending stage to process the data, including steps:

S2.1: The data sending end inserts training symbols into the data symbols. The insertion method is: take N sending symbols as a group, and then divide the N sending symbols into R groups. Each group of sending symbols contains One training symbol, the sequence number of R training symbols in the parallel symbol is recorded as, i≤r≤w;

S2.2: Send the data signal to the optical burst receiver, perform serial-to-parallel conversion on the data signal that has been coherently demodulated and sampled and quantized to obtain N parallel signals. The first group of parallel signals enters N parallel signal processing branches, and the data The parallel signal processing branch in the transmission stage includes an equalizer, a frequency offset estimation module and a decision module. The equalizer in the first branch processes the equalized signal.

S2.3: Perform frequency offset estimation on N branches respectively:

When the branch is a training signal, the frequency offset of the equalized signal ^ is estimated directly based on the known training symbol _e , and the cumulative phase error ¾ and frequency offset estimate Δ „ are obtained; when the branch is a data signal, the equalized signal is first ¾ Perform phase compensation, and the signal after phase compensation ¾ is:

Among them, ^ represents the delay of the average frequency offset estimate, "·1 represents rounding up; the decision module determines the signal to obtain the decision signal, and estimates the frequency offset of the equalized signal ¾ according to the decision signal to obtain the frequency offset estimate Δ „ and Cumulative phase error <¾;

S2.4: Average the estimated frequency offset values Δώ„ of N branches to obtain the estimated average frequency offset ΔΓ of the nth group of parallel signals;

S2.5: N branches calculate their error signals respectively

When the branch is a training signal, g^' =, the error signal is:

^£ n,i = ^m k —

When the branch is a data signal, the error signal^ is:

ε _η . = m _k -e ^A · e ^J Ά ¹

S2.6: Update the equalizer tap system used by the M + l group of parallel symbols

^ = _n -- -∑[s _n _ _DC -V(nD/ _c )

c 2 1

Among them, Γ is the iteration step length set in the data sending stage, N: represents the number of error signals selected from N branches to participate in the tap coefficient calculation in the data sending stage, i≤N:≤N,

S2.7: Determine whether the data signal has been processed. If not, return to step S2.2 to continue processing. If the processing is completed, it ends. 2. The joint parallel method according to claim 1, characterized in that the specific method of frequency offset estimation includes the following steps:

S3.1: Calculate the cumulative phase error of the equalized signal ¾:

When the branch is a training symbol, the cumulative phase error ¾=ΦΑ_, where represents the phase of the known training symbol and represents the phase of the equalized signal ¾;

When the branch is a data signal, the cumulative phase error <¾ =3^_, where 4 represents the phase of the data decision signal e ⁱ ;

S3.

2: Calculate the estimated frequency offset of this branch ΔώΤ^^ - ^ where represents the cumulative phase error of the A-1th signal.

3. The parallel method according to claim 1, characterized in that, the insertion position of the training symbol in step S2.1 is N/(Rx2) +; cxN/R, = 0, l,..., R.