WO2015103805A1

WO2015103805A1 - Frequency-domain dispersion search method based on training sequence

Info

Publication number: WO2015103805A1
Application number: PCT/CN2014/072121
Authority: WO
Inventors: 於涛; 许渤; 邱昆
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: CN103746950B; CN103746950A

Abstract

A frequency-domain dispersion search method based on a training sequence, comprising performing a frequency-domain dispersion search using a modulation symbol sequence obtained by modulating a training sequence. A shift factor Ω_CT = N_FFT/2^K+2 is used when a receiving end performs a dispersion search based on a frequency-domain maximum cost function on a continuous sampling signal which was received, where K is an integer greater than or equal to 0; and at the same time, the following preprocessing is correspondingly performed on a modulation symbol sequence of a sending end: taking 2^Κ modulation symbol sequences as a group, and setting all the modulation symbols as zero at each interval of one group, with the others remaining unchanged. Thus, the problem that the amplitude of a frequency-domain maximum cost function changes with an ADC sampling point moment is solved, and a larger chromatic dispersion search step length may be allowed to be used to perform a chromatic dispersion search, thereby reducing the number of searches, lowering the complexity of the algorithm and saving system resources.

Description

FIELD OF THE INVENTION The present invention relates to the field of frequency domain dispersion search technology, and more particularly to a frequency domain dispersion search method based on training sequences. BACKGROUND OF THE INVENTION With the development of fiber coherent communication technology, a coherent detection receiver including digital equalization technology can compensate for all linear channel loss or low nonlinear fiber channel, which makes it possible to color at the receiving end during transmission. Degree dispersion (CD, Chromatic Dispersion) performs digital equalization without compensating in the optical domain by Dispersion Compensating Fiber (DCF). In the commonly used digital equalization methods, there are two ways of domain equalization and frequency domain equalization. Due to the very low computational complexity and good parallelism of the frequency domain equalization, the Finite Impulse Response (FIR) equalizer and the Least Mean Square equalizer (LMS) in time domain. In contrast, the frequency domain equalizer has become the industry's main focus on the chromatic dispersion balance technology.

In the frequency domain based method, the time domain signal after sampling at the receiving end is first transformed into a frequency domain by a fast Fourier transform (FFT) and multiplied by a frequency domain dispersion compensation function, which can be regarded as a fiber channel. The inverse function of the transmitted dispersion function, and finally the compensated frequency domain signal is transformed into a continuous data stream by inverse fast Fourier transform (IFFT) transformation to the time domain. The disadvantage of this method is that the frequency domain dispersion compensation function must be utilized on the basis of knowing the channel dispersion size in advance. However, since the dispersion value of the transmission link is variable in a dynamic optical network environment, the amount of dispersion in the channel needs to be estimated first.

At present, a commonly used dispersion estimation method is a frequency domain dispersion search algorithm, which uses a best matching search method to observe the parameters of a set of dispersion filters, which is based on the cost function of calculating the signals after each group of equalizations. Make a judgment to achieve.

1 is a schematic structural diagram of a fiber coherent communication receiver based on frequency domain dispersion search. As shown in FIG. 1, the received polarization multiplexed signal is split into two polarization directions by a polarization beam splitter for independent coherent demodulation. The four-channel electrical signal after coherent demodulation first needs to perform double-rate AD sampling and quantization, and then the sampled signal is transformed into the frequency domain by FFT and multiplied by the frequency domain dispersion compensation function to transmit the frequency domain dispersion compensation function according to the transmission. The range of channel dispersion is divided into groups to satisfy different dispersion searches For the long-term needs, a dispersion search is performed based on the compensated frequency domain signal, and the dispersion range of the channel transmission is determined. Finally, the set of frequency domain signals with the best compensation effect is selected to be processed by IFFT to the time domain for subsequent processing.

The dispersion search algorithm based on the frequency domain maximum cost function is a commonly used method for searching and estimating the dispersion in the channel. The key of the dispersion search algorithm is to calculate the frequency domain maximum cost function based on the received sampled signals for several possible values of the channel dispersion, and then select the channel dispersion value corresponding to the largest frequency domain cost function value as the result of the dispersion search and estimation. . Figure 2 is a schematic diagram of a dispersion search algorithm based on the frequency domain maximum cost function. As shown in Figure 2, the dispersion-based search algorithm based on the frequency domain maximum cost function performs chromatic dispersion estimation by combining the signals in the polarization directions of X and Y. The signal processing method in a single polarization direction is:

First, the received continuous sampling signal is divided into N groups according to the number of FFT conversion points.

= l,2,-,N , transform each set of signals into the frequency domain using FFT to generate a digital spectrum of R _p [M], M = _N _FFr /2, ---, N _FFr /2_l. Secondly, according to the channel dispersion range of the transmission, the frequency domain dispersion compensation function H _cz M = _eX p (2 Δ/) ί = \, 2, ····, Μ where =U _min + is defined for the M possible transmission channel dispersion values. AU _mm + ² AL, —, Z _max represents the number of transmission distance values selected during the dispersion search,

ΔΙ represents the resolution of the dispersion search, ^ represents the second-order dispersion effect coefficient, and Δ/ represents the frequency domain sampling interval Af = RN _FFT , which represents the sampling rate. Then, the frequency domain dispersion compensation function can be-signal-set corresponding to the M possible transmission channel dispersion values in the frequency domain _{_{compensation,; · W = H CD [}} n] -R p [n], in accordance with the compensation. The latter signal can define the domain cost function c^[Q _CT ]= £(R [M]0+n _CT D, where .^ denotes the shift factor, superscript

«二 1

* indicates complex conjugate. The frequency domain cost function is related to clock recovery, which is represented by the phase and frequency information of the clock frequency obtained from the autocorrelation function of the complex signal after frequency domain compensation. Finally, the frequency domain maximum cost function corresponding to each possible transmission channel dispersion value is defined as: where |(^·[Ω^]| represents the magnitude of C _cr ].

The influence of the search, as shown in Figure 2, is based on the frequency domain maximum cost function of the dispersion search algorithm, which is X,

Y is the average of the most expensive price in the frequency domain corresponding to the direction of polarization, ie,

J _FD ] = ]| The maximum value of the M frequency domain maximum costs corresponds to

The transmission channel dispersion value is the estimated channel dispersion.

However, in the above-mentioned dispersion-based search algorithm based on the frequency domain maximum cost function, there are two problems: First, under the most commonly used double rate sampling, n _CT = N^ _r / ^{2 is used} , but in this case Next, the amplitude of the frequency domain maximum cost function will change with the time of the sampling point of the ADC, resulting in a large performance difference, and even causing the dispersion search module to not work properly. Secondly, in the dispersion search algorithm, the number of dispersion values that need to be tried determines the complexity of the dispersion search algorithm. In the traditional dispersion search algorithm, it is necessary to use a small dispersion search length ΔΙ to ensure that the correct dispersion estimation value can be obtained. In the actual processing, multiple sets of frequency domain dispersion compensation functions are needed to achieve the required dispersion range. Search, this is a waste of resources in hardware implementation. Summary of the invention

The object of the present invention is to overcome the deficiencies of the prior art, and to provide a frequency domain dispersion search method based on a training sequence, which adjusts the 0^ shift factor and the modulation symbol of the training sequence, which can not only solve the influence of the ADC sampling time, but also allow Use a larger dispersion search to reduce the complexity of the dispersion search.

To achieve the above object, the present invention is based on a frequency domain dispersion search method of a training sequence, comprising the following steps:

S1: At the transmitting end, preprocessing the modulated symbol sequence obtained by the training sequence, and the preprocessing method is: grouping the modulation symbol sequence into a group, which is an integer greater than or equal to 0, and the modulation symbol is separated by 1 group. Set to 0, the rest remains unchanged;

S2: The digital coherent optical receiver performs coherent demodulation on the received signal to obtain four electrical signals, and respectively performs two-rate sampling and quantization on each electrical signal;

S3: performing a dispersion search based on a frequency domain maximum cost function on the continuous sampled signal in step S2 to obtain a dispersion range of the estimated channel, wherein the shift factor Q _CT = N _F ^/2^ ² .

The invention is based on a frequency domain dispersion search method of a training sequence, and uses a modulation symbol sequence obtained by modulating a training sequence to perform frequency domain dispersion search, and the receiving end performs frequency-based transmission on the received continuous sampling signal. The dispersion search of the domain maximum cost function uses different shift factors Q _CT = N _F ^/2^ ² , K is an integer greater than or equal to 0, and the corresponding modulation symbol sequence at the transmitting end is preprocessed as follows: The symbol sequence is grouped by ^, and the modulation symbol is set to 0 every interval of 1 group, and the rest remains unchanged. This solves the problem that the amplitude of the frequency domain maximum cost function varies with the ADC sampling point time. It allows the dispersion search to be performed with a larger dispersion search length, thereby reducing the number of searches, reducing the complexity of the algorithm, and saving system resources. DRAWINGS

FIG. 1 is a schematic structural diagram of a fiber coherent communication receiver based on frequency domain dispersion search;

2 is a schematic diagram of a dispersion search algorithm based on a frequency domain maximum cost function;

3 is a schematic diagram of preprocessing results of a modulation symbol sequence;

4 is a comparison diagram of dispersion search lengths corresponding to four shift factors;

Figure 5 is a comparison of amplitude variation of the frequency domain maximum cost function for different ADC sampling moments; Figure 6 is a simulation result of four channel dispersions in the range of 200 km;

Figure 7 is a distribution diagram of 5000 transmission channel dispersion and estimated channel dispersion;

Figure 8 is a simulation result of four channels of dispersion in a range of 500 km;

Figure 9 is a simulation result of four channel dispersions within the adjusted 500 km range. detailed description

The specific embodiments of the present invention are described below in conjunction with the drawings in order to provide a better understanding of the invention. 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

The invention is based on the frequency domain dispersion search method of the training sequence, mainly achieves the object of the invention by adjusting the training sequence modulation symbol and the shift factor, and the specific steps include:

S101: At the transmitting end, preprocessing the modulated symbol sequence obtained by the training sequence, and the preprocessing method is: grouping the modulation symbol sequence into a group, which is an integer greater than or equal to 0, and the modulation symbol is separated by one group. Set to 0, the rest remains unchanged.

3 is a schematic diagram of preprocessing results of a sequence of modulation symbols. As shown in Figure 3, for the same foundation The modulation symbols, when equal to 0, 1, and 2, respectively, comprise a set of 1, 2, and 4 modulation symbols, respectively, and the modulation symbols are all set to 0 for each interval. This preprocessing, together with the subsequent shift factor change, can solve the problem that the amplitude of the frequency domain maximum cost function changes with the time of the ADC sampling point, and can also suppress the data after sampling from canceling each other in phase.

S102: The digital coherent optical receiver performs coherent demodulation on the received signal to obtain four electrical signals, and performs double-rate sampling and quantization on each electrical signal.

S103: performing a dispersion search based on a frequency domain maximum cost function on the continuous sampling signal in step S2 to obtain a dispersion range of the estimated channel, where the shift factor Q _CT = N _F ^/2^ ² . It can be seen that when respectively equal to 0, 1, 2, the corresponding shift factor. ^ are N _ffr / ⁴ and UN^/16, respectively. With the shift factor of the present invention, the dispersion search can be realized with a larger dispersion search length ΔΙ relative to the shift factor of A^ _r / ² in the prior art.

After obtaining the dispersion range of the estimated channel, the estimated channel dispersion can be directly used for dispersion compensation during the actual data transmission.

FIG. 4 is a comparison diagram of dispersion search lengths corresponding to four shift factors. As shown in FIG. 4, four shift factors N _ffr /2, N _FFT , N _FFT I &, N _FFT /\6. The simulation is implemented in a 112 Gb/s PDM-QPSK system in which the transmission symbol rate is 28 GBaud/s and the channel dispersion is set to 100 km, corresponding to the peak point of the maximum cost function of each frequency domain in FIG. It can be seen that when the shift factor Q _CT = N^ _r / ² , the actual dispersion search length that can be used is small. It is necessary to calculate a large number of frequency domain maximum cost functions to determine the range of channel dispersion. With the other three shift factors, a larger dispersion search length can be used. For example, when Q _CT = N _ffr /16, the search for the dispersion can be as long as 120 km. Of course, the smaller the dispersion search length, the more accurate the search results will be. However, in the case of large limitations on algorithm complexity and hardware resources, the present invention can allow a larger dispersion search length to achieve a more accurate dispersion search without biasing the estimated channel dispersion and the transmission channel dispersion. Big.

Figure 5 is a comparison of amplitude variation of the frequency domain maximum cost function for different ADC sampling moments. In the comparative simulation of Fig. 5, when the shift factor n _CT = N _ffr / ² , the modulation symbol sequence obtained by the training sequence is not preprocessed, and the frequency domain maximum cost function can be seen at different ADC sampling moments. The amplitude variation varies greatly, and the maximum amplitude is not necessarily obtained even in the case of complete compensation without residual dispersion. When the shift factor Q _CT = N _ffr /16, the modulation symbols obtained by the training sequence are preprocessed, and the amplitudes of the cost _parameters of the frequency domain are almost different for different ADC sampling moments. No effect.

The dispersion search simulation of the present invention is performed by taking the shift factor Q _CT = N _ffr / 16 as an example. For n _cT = N _FFT /l6 , only four frequency domain dispersion compensation functions are needed at this time to search for arbitrary dispersion values of channel dispersion in the range of 200 km by dispersion search based on the frequency domain maximum cost function. Set 匪 = 25 km, ΔΙ = 50 km, at this time Α = 25, ₂ = 75, ₃ = 125, L ₄ = \75, ie compensate 25 km, 75 km, 125 km and 175 km respectively. Finally, the four sets of compensated frequency domain signals are respectively determined by the obtained maximum cost function amplitude to determine the range of estimated channel dispersion, that is, which frequency domain maximum cost function has the largest amplitude, and the determined dispersion range is Estimated channel transmission dispersion range.

In the 112Gb/s PDM-QPSK modulation system, the second-order dispersion coefficient is -20ps ^A 2/km, and the optical signal-to-noise ratio is

12dB, the steady-state frequency offset of the laser is set to 3G and the transient frequency offset is added. In the transmission, 4096 modulation symbols obtained by training sequence modulation are used for simulation, and the receiving end samples each symbol twice. A total of 8192 sampling points are divided into 8 groups, and each group of 1024 points is subjected to FFT and 4 channel transmission distance values are respectively calculated. The frequency domain cost function, the final calculated four frequency domain maximum cost function magnitudes and the decision is made.

Figure 6 is a simulation result of four channel dispersions in the range of 200 km. As shown in Fig. 6, the actual channel dispersion corresponds to transmission distance values of 10, 50, 125, and 200 kilometers, respectively. The four values in each curve are the frequency domain maximum cost function magnitudes of the channel dispersions that compensate for 25, 75, 125, and 175 kilometers, respectively. When the actual transmission distance value is 10 km, among the four frequency domain maximum cost function amplitudes, the frequency domain maximum cost function of 25 km corresponds to the largest amplitude, so the estimated transmission distance value is within 25 km. Since the dispersion of the actual transmission distance value of 50 km is located at the midpoint of the compensated 25 and 75 km, the frequency domain maximum cost function amplitude corresponding to 25 km and 75 km does not differ much, and the determined channel transmission dispersion is 25 km or It is accurate within 75 km. The actual transmission distance value of 125 km is exactly at the compensated 125 km, and the maximum frequency domain maximum cost function can be obtained. When the actual transmission distance value is 200 km, the frequency domain maximum cost function corresponding to 175 km is the largest, so the estimated transmission distance value is within 175 km.

In order to make the simulation results more general, in the simulation, 5000 different transmission channel dispersions are randomly generated in the range of 200 km dispersion, and the simulation conditions are the same as above. Figure 7 is a plot of 5000 transmission channel dispersion and estimated channel dispersion. As shown in Fig. 7, the reason why the distribution of the transmission channel dispersion and the estimated channel dispersion are slightly different is that the dispersion of the transmission is compensated after being compensated at 50, 100, and 150 km. Either side is correct, such as transmitting 50 km of dispersion, which is the same as the compensation of 25 km and 75 km. The result of the search is 25 km or 75 km.

In addition, the dispersion search within a range of 500 km was simulated. Figure 8 is a graph showing simulation results for four channel dispersions over a range of 500 kilometers. Similarly, the actual channel dispersion corresponding transmission distance values in the simulation are 10, 250, 312.5, and 500 kilometers, respectively. The dispersion search length ΔΙ = 125 km of the frequency domain dispersion compensation function is set, and the compensation is 62.5 km and 187.5 km, respectively. 312.5 km and 437.5 km, shift factor Q _CT = N^ _r /16. It is found by simulation that when Q _CT = N^ _r /16, although an accurate search can be made, only when the transmission channel dispersion is close to the compensated channel dispersion, the ratio of the maximum value to the minimum value of the frequency domain maximum cost function is more Conducive to the judgment. Therefore, the shift factor is changed further. The size of ^, ie, = 3, makes. _Cr = W _FFr /32, correspondingly, the transmitted training sequence modulation symbols shall be grouped with 8 modulation symbols, each set is set to 0, and the other modulation symbols remain unchanged. Figure 9 is a simulation result of four channels of dispersion in the adjusted 500 km range. As shown in Fig. 9, compared with Fig. 8, it can be seen that the difference between the amplitudes of the four frequency domain maximum cost functions is more obvious, which is more conducive to making an accurate search. It can be seen that as the search range increases, the value can be increased accordingly. At present, the commonly used digital coherent optical communication transmission distance values are generally in the range of tens to hundreds of kilometers, so the commonly used values are 0, 1, 2, 3.

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 frequency domain dispersion search method based on training sequence, which is characterized by including the following steps: S1: At the transmitting end, preprocess the sequence modulation symbol sequence obtained by modulating the training sequence. The preprocessing method is: The modulation symbol sequence is grouped into a group, and is an integer greater than or equal to 0. The modulation symbols are all set to 0 for every interval group, and the rest remain unchanged;

S2: The digital coherent optical receiver performs coherent demodulation on the received signal to obtain 4 electrical signals, and each electrical signal is sampled and quantized at twice the rate;

S3: Perform a dispersion search based on the maximum value cost function in the frequency domain for the continuously sampled signal in step S2 to obtain the estimated channel dispersion range, where the shift factor Q _CT = N _F ^/2^ ² .

2. The frequency domain search method according to claim 1, characterized in that the value range of is 0≤≤3.