CN101895499B - Method for demultiplexing polarization by using constant rotation sign training sequence - Google Patents

Abstract

The invention belongs to the field of optical communication, and is applied to the polarization demultiplexing of a polarization diversity multiplexing coherent optical communication system. According to the invention, for the problem of the polarization demultiplexing in the polarization diversity multiplexing coherent optical communication system, a special training sequence and related algorithms are designed. Through the method, a coherent receiver can fast, automatically and accurately perform the polarization demultiplexing. As the track of the training sequence, viewed from a planisphere, rotates along a unit circle after sign power is normalized, the training sequence provided by the invention is called a constant rotation sign training sequence. The invention discloses two processing algorithms for the constant rotation sign training sequence, which are a frequency-domain algorithm and a time-domain algorithm respectively, wherein the frequency-domain algorithm can be combined with frequency-domain dispersion compensation without adding a frequency spectrum calculation module. Through the method, the receiver does not need synchronizing with an initial position of the training sequence.

Description

By the multiplexing method of constant rotation sign training sequence depolarization

Technical field

The invention belongs to optical communication field, be applied to the multiplexing coherent optical communication system of polarization diversity, carry out polarization demultiplexing.

Background technology

Coherent optical communication not only can improve spectrum efficiency with high order modulation technology, polarization diversity multiplex technique, can also use self-adaptive digital signal treatment technology to carry out the damage of dynamic compensation signal.It is a kind of important technology that has broad prospect of application.In the multiplexing coherent optical communication system of polarization diversity, transmitting terminal is input to orthogonal polarization direction two ways of optical signals in optical fiber and transmits.Fig. 1 represents the input signal of polarization diversity multiplex system.Light signal transmission in optical fiber can be subject to the impact of optic fiber polarization mould dispersion.Polarization mode dispersion meeting is rotated the polarization state of light signal.As shown in Figure 2.And optic fiber polarization mould dispersion is random, causes through the polarization state of the light signal of Optical Fiber Transmission and rotate at random.The light signal polarization state of receiving terminal as shown in Figure 3.

As shown in Figure 4, reception signal 401 is resolved into two mutually perpendicular light signals of polarization state by polarization beam apparatus 402 to the multiplexing coherent optical communication system receiving terminal of polarization diversity, enters respectively X polarization branch 403 and Y polarization branch 404.Identical with optical signal carrier frequency or the very approaching laser of local laser (LO) 405 transmitting, is divided into two-way, enters two 90 ° of frequency mixers 406,407 respectively with the signal of X polarization branch 403, Y polarization branch 404.The output of two 90 ° of frequency mixers is received and changes into the signal of telecommunication by photodiode 408, and the signal of telecommunication is by after analog-to-digital conversion device (ADC) 409 samplings, and what obtain is the baseband digital signal 410 of receiving optical signals 401.Baseband digital signal enters DSP module 411 and carries out the operations such as dispersion compensation, clock synchronous, polarization demultiplexing, polarization mode dispersion (PMD) compensation, demodulation, recovers initial data 412.

Polarization beam apparatus 402 placing directions of receiving terminal are fixed, and on fixing orthogonal two polarization directions, receive signal.Under normal circumstances, two polarization directions of the light signal that arrives receiving terminal can not be aimed at respectively in these two polarization directions.This makes each some X polarization branch 403 and the Y polarization branch 404 that enters receiving terminal of two light signals on polarization direction.Reception signal in X polarization branch 403 and Y polarization branch 404 is the stack of two transmitted signals on polarization direction.Conventionally coherent receiver (Fig. 4) uses the butterfly filter 506 in Digital Signal Processing (DSP) module (Fig. 5) to carry out polarization demultiplexing, and giving the rational numerical value of butterfly filter configuration is the multiplexing key of depolarization.

The present invention is intended to solve the polarization demultiplexing problem of the multiplexing coherent optical communication system of polarization diversity.

Summary of the invention

The present invention is applicable in the multiplexing coherent optical communication system of polarization diversity, and the modulation system of system can be MPSK (M >=4) or MQAM (M >=4).The present invention is directed to the problem of polarization demultiplexing in the multiplexing coherent optical communication system of polarization diversity, designed a kind of special training sequence and related algorithm.Transport coherent receiver in this way and can carry out fast and automatically, exactly polarization demultiplexing.The training sequence that the present invention proposes, after symbol power normalization, rotates on unit circle at its track of planisphere, therefore called after constant rotation sign training sequence.On two polarization directions of transmitting terminal, send respectively the identical and inconstant symbol sebolic addressing of velocity of rotation of certain length, the implication that wherein velocity of rotation is not identical comprises that slewing rate is not identical or rotation direction is different (for example, on a polarization direction, turn clockwise, on another polarization direction, be rotated counterclockwise).The velocity of rotation of the training sequence that the known transmitting terminal of coherent receiver sends on two polarization directions.The present invention proposes two kinds of Processing Algorithm for constant rotation sign training sequence.These two kinds of algorithms, for receiving terminal, give suitable weights can to butterfly filter 506, carry out polarization demultiplexing.

In these two kinds of algorithms, a kind of algorithm need to calculate the frequency spectrum that receives signal, estimates channel transfer functions matrix from the frequency domain that receives signal, thereby gives butterfly filter 506 initializes.Therefore, this algorithm called after frequency domain algorithm.The key step of algorithm comprises:

(1) ask discrete Fourier transform (DFT) to calculate frequency spectrum to the training sequence receiving;

(2) from frequency spectrum, find peak value;

(3) go out channel transfer functions from peak estimation;

(4) with transfer function contrary come the weights of initialization butterfly filter.

The position occurring from spectrum peak can also estimate local oscillator coherent receiver (LO) 405 and the frequency difference that sends optical signal carrier easily.

Another kind of algorithm belongs to least mean-square error (LMS) algorithm, in the time domain that receives signal, processes, and optimizes the weights of butterfly filter 506.Therefore, this algorithm called after Time-Domain algorithm.The key step of algorithm comprises:

(1) angle of upper moment butterfly filter two branch output signals of calculating (this signal is plural number);

(2) extrapolate the desired value of current time output signal from the angle of a upper moment output signal and the training sequence symbols velocity of rotation of this polarization branch;

(3) difference of the actual value of the desired value of current time output signal and current time output signal is as the error signal of LMS algorithm.

Brief description of the drawings

Fig. 1: the schematic diagram that transmits of polarization diversity multiplex system;

Fig. 2: optic fiber polarization mould dispersion makes the schematic diagram of the light signal polarization state Random-Rotation of transmission;

Fig. 3: receiving terminal is receiving optical signals polarization state schematic diagram just;

Fig. 4: the receiver of the multiplexing coherent optical communication system of polarization diversity;

Fig. 5: the DSP module in the receiver of the multiplexing coherent optical communication system of polarization diversity;

Fig. 6: the planisphere of the multiplexing QPSK system of polarization diversity, the rotation schematic diagram of constant rotation sign training sequence on planisphere;

Fig. 7: the frequency domain algorithm structure chart that carries out polarization demultiplexing with constant rotation sign training sequence;

Fig. 8: frequency domain algorithm detail flowchart;

Fig. 9: the Time-Domain algorithm structure chart that carries out polarization demultiplexing with constant rotation sign training sequence;

Figure 10: Time-Domain algorithm detail flowchart.

Embodiment

As shown in Figure 4, coherent receiver comprises front-end processing part 413 and Digital Signal Processing (DSP) module 411 to the receiver of polarization diversity multipurpose photo-communication system.Front-end processing part 413 becomes light signal into baseband digital signal 410.DSP module 411 is processed baseband digital signal 410, carries out clock synchronous, compensates various signal impairments, and the operation recoveries such as demodulation go out data 412.The detailed structure of DSP module 411 as shown in Figure 5,412 in the corresponding coherent receiver of output 510 of DSP module, the input 501,502 of DSP module connects the baseband digital signal 410 in coherent receivers.Notice the input x of DSP circuit _in501, y _inthe 502nd, plural number, with 410 relation is:

x _in＝x _Re+i×x _Im

y _in＝y _Re+i×y _Im

Input signal x _in501, y _infirst 502 enter 503 carries out dispersion compensation and clock synchronous.The output x of dispersion compensation and clock synchronization module 503 _p504, y _pthe 505th, synchronous signal, then enters butterfly filter 506 and carries out polarization demultiplexing, polarization mode dispersion (PMD) compensation and residual chromatic dispersion compensation.Butterfly filter 506 is made up of 4 complex filters 511,512,513,514 and two adders 515,516.The weights h of complex filter 511,512,513,514 _xx, h _xy, h _yx, h _yybeing plural number, can be one or more dimensions vector.Being operating as of butterfly filter 506:

x _out＝x _p*h _xx+y _p*h _xy

y _out＝x _p*h _xy+y _p*h _yy

Wherein, the variable in formula is all plural number; * be convolution operation; x _out507, y _outthe 508th, filter output; Suitably set the weights h of filter _xx, h _xy, h _yx, h _yyget final product polarization demultiplexing.The output 507,508 of butterfly filter enters phase bit recovery reconciliation mode transfer piece 509 and obtains demodulating data 510.

The linear channel model of polarization diversity multipurpose photo-communication system

The transmitted signal of polarization diversity multiplex system, time domain and frequency domain representation are respectively:

$\stackrel{\→}{s}\left(t\right)=\left[\begin{array}{c}{s}_x\left(t\right)\\ s_y\left(t\right)\end{array}\right],$ $\stackrel{\→}{S}\left(\mathrm{\ω}\right)=\left[\begin{array}{c}{S}_x\left(\mathrm{\ω}\right)\\ S_y\left(\mathrm{\ω}\right)\end{array}\right]$

Wherein x, y is two orthogonal polarization directions.

The time domain and the frequency domain representation that receive signal are respectively:

$\stackrel{\→}{r}\left(t\right)=\left[\begin{array}{c}{r}_x\left(t\right)\\ r_y\left(t\right)\end{array}\right],$ $\stackrel{\→}{R}\left(\mathrm{\ω}\right)=\left[\begin{array}{c}{R}_x\left(\mathrm{\ω}\right)\\ R_y\left(\mathrm{\ω}\right)\end{array}\right]$

Transmission matrix:

$\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left(\mathrm{\ω}\right)& H_{\mathrm{xy}}\left(\mathrm{\ω}\right)\\ H_{\mathrm{yx}}\left(\mathrm{\ω}\right)& H_{\mathrm{yy}}\left(\mathrm{\ω}\right)\end{array}\right]$

Transmission equation:

$\left[\begin{array}{c}{R}_x\left(\mathrm{\ω}\right)\\ R_y\left(\mathrm{\ω}\right)\end{array}\right]=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left(\mathrm{\ω}\right)& H_{\mathrm{xy}}\left(\mathrm{\ω}\right)\\ H_{\mathrm{yx}}\left(\mathrm{\ω}\right)& H_{\mathrm{yy}}\left(\mathrm{\ω}\right)\end{array}\right]\·\left[\begin{array}{c}{S}_x\left(\mathrm{\ω}\right)\\ S_y\left(\mathrm{\ω}\right)\end{array}\right]$

Constant rotation sign training sequence

The training sequence that the present invention proposes, after symbol power normalization, rotates on unit circle at its track of planisphere, therefore called after constant rotation sign training sequence.On two polarization directions of transmitting terminal, send respectively the identical and inconstant symbol sebolic addressing of velocity of rotation of certain length, the implication that wherein velocity of rotation is not identical comprises that slewing rate is not identical or rotation direction is different (for example, on a polarization direction, turn clockwise, on another polarization direction, be rotated counterclockwise).

On X polarization signal on planisphere with angular frequency _xrotate:

s _x(t)＝exp(iφ _x)exp(iω _xt)

Wherein, φ _xthe initial phase of training sequence, t=0,1,2...

On Y polarization signal on planisphere with angular frequency _yrotate:

s _y(t)＝exp(iφ _y)exp(iω _yt)

Wherein, φ _ythe initial phase of training sequence, t=0,1,2...

Note, X, the velocity of rotation on Y polarization direction is not identical, i.e. ω _x≠ ω _y.

For example,, for QPSK signal, ω _x, ω _ybe taken as respectively the rotation direction of training sequence on planisphere as shown in Figure 6.On X polarization direction, send the planisphere of training sequence as shown in Figure 60 5, the phase place of symbol sebolic addressing is followed successively by deng, by that analogy.On Y polarization direction, send the planisphere of training sequence as shown in Figure 61 0, the phase place of symbol sebolic addressing is followed successively by deng, by that analogy.

X polarization direction sends the frequency domain representation of upper signal:

$S_x\left(\mathrm{\ω}\right)=\left\{\begin{array}{cc}\exp \left({\mathrm{i\φ}}_x\right),& \mathrm{\ω}={\mathrm{\ω}}_x\\ 0,& \mathrm{\ω}\≠{\mathrm{\ω}}_x\end{array}\right.$

The frequency domain representation of transmitted signal on Y polarization direction:

$S_y\left(\mathrm{\ω}\right)=\left\{\begin{array}{cc}\exp \left(i{\mathrm{\φ}}_y\right),& \mathrm{\ω}={\mathrm{\ω}}_y\\ 0,& \mathrm{\ω}\≠{\mathrm{\ω}}_y\end{array}\right.$

By S _x(ω), S _y(ω) expression formula is brought side signal transmission journey into:

$\left[\begin{array}{c}{R}_x\left(\mathrm{\ω}\right)\\ R_y\left(\mathrm{\ω}\right)\end{array}\right]=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left(\mathrm{\ω}\right)& H_{\mathrm{xy}}\left(\mathrm{\ω}\right)\\ H_{\mathrm{yx}}\left(\mathrm{\ω}\right)& H_{\mathrm{yy}}\left(\mathrm{\ω}\right)\end{array}\right]\·\left[\begin{array}{c}{S}_x\left(\mathrm{\ω}\right)\\ S_y\left(\mathrm{\ω}\right)\end{array}\right]$

Obtain:

R _x(ω)＝H _xx(ω _x)exp(iφ _x)+H _xy(ω _y)exp(iφ _y)

R _y(ω)＝H _yx(ω _x)exp(iφ _x)+H _yy(ω _y)exp(iφ _y)

Carry out the frequency domain algorithm of polarization demultiplexing with constant rotation sign training sequence:

Fig. 7 is the structure chart of frequency domain algorithm, and transmitting terminal sends constant rotation sign training sequence, and the input that arrives butterfly filter is x _p701, y _p702.Butterfly filter is output as x _out705, y _out706.X _p701, y _p702 enter channel transfer functions estimator 704, and estimator 704 estimates channel transfer functions and the weights initialize to butterfly filter 703 according to transfer function.Butterfly filter 703 after assignment can successfully carry out polarization demultiplexing.Fig. 8 is the flow chart of transfer function estimator 704.803,804 couples of input signal x of discrete Fourier transform (DFT) module _p801, y _p802 ask frequency spectrum, and frequency spectrum is R _x(ω) 805 and R _y(ω) 806.Frequency spectrum R _x(ω) 805 and R _y(ω) 806 be input to respectively and get peak value module 807,808.807,808 couples of R of peak value module _x(ω) and R _y(ω) at ω=ω _xand ω=ω _ynear the point of finding absolute value maximum is peak value, altogether should find 4 peak values, is respectively:

1.R _x(ω) at ω=ω _xnear peak value size is h ₁=H _xx(ω _x) exp (i φ _x), peak is ω=ω ₁;

2.R _x(ω) at ω=ω _ynear peak value size is h ₂=H _xy(ω _y) exp (i φ _y), peak is ω=ω ₂;

3.R _y(ω) at ω=ω _xnear peak value size is h ₃=H _yx(ω _x) exp (i φ _x), peak is ω=ω ₃;

4.R _y(ω) at ω=ω _ynear peak value size is h ₄=H _yy(ω _y) exp (i φ _y), peak is ω=ω ₄.

Peak value size 812,813 is input to for estimating channel transfer functions matrix 814, and peak 809,810 can estimate the frequency difference 811 of local oscillator (LO) 405 and flashlight carrier wave.814 modules are estimated channel transfer functions matrix from spectrum peak.Concrete operations are: frequency spectrum R _x(ω) and R _y(ω) peak value in composition matrix H ':

$H^{\′}=\left[\begin{array}{cc}{h}_1& h_2\\ h_3& h_4\end{array}\right]=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left({\mathrm{\ω}}_x\right)\exp \left({\mathrm{i\φ}}_x\right)& H_{\mathrm{xy}}\left({\mathrm{\ω}}_y\right)\exp \left({\mathrm{i\φ}}_y\right)\\ H_{\mathrm{yx}}\left({\mathrm{\ω}}_x\right)\exp \left({\mathrm{i\φ}}_x\right)& H_{\mathrm{yy}}\left({\mathrm{\ω}}_y\right)\exp \left({\mathrm{i\φ}}_y\right)\end{array}\right]$

H ' can be written as:

$H^{\′}=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{xy}}\left({\mathrm{\ω}}_y\right)\\ H_{\mathrm{yx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{yy}}\left({\mathrm{\ω}}_y\right)\end{array}\right]\·\left[\begin{array}{cc}\exp \left(i{\mathrm{\φ}}_i\right)& 0\\ 0& \exp \left({\mathrm{i\φ}}_y\right)\end{array}\right]=H\·\mathrm{\Φ}$

Wherein, $H=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{xy}}\left({\mathrm{\ω}}_y\right)\\ H_{\mathrm{yx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{yy}}\left({\mathrm{\ω}}_y\right)\end{array}\right],$ $\mathrm{\Φ}=\left[\begin{array}{cc}\exp \left({\mathrm{i\φ}}_x\right)& 0\\ 0& \exp \left({\mathrm{i\φ}}_y\right)\end{array}\right].$

The length of supposing the unit impulse response of channel transfer functions is 1.Like this, its Fourier transform H _xx(ω), H _xy(ω), H _yx(ω), H _yy(ω) on all ω, equate.So can estimate channel transfer functions:

$\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left(\mathrm{\ω}\right)& H_{\mathrm{xy}}\left(\mathrm{\ω}\right)\\ H_{\mathrm{yx}}\left(\mathrm{\ω}\right)& H_{\mathrm{yy}}\left(\mathrm{\ω}\right)\end{array}\right]=\left[\begin{array}{cc}{H}_{\mathrm{xx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{xy}}\left({\mathrm{\ω}}_y\right)\\ H_{\mathrm{yx}}\left({\mathrm{\ω}}_x\right)& H_{\mathrm{yy}}\left({\mathrm{\ω}}_y\right)\end{array}\right]=H$

To the H ' H ' that inverts ^-1, and with H ' ^-1compose weights to butterfly filter.The total transfer function of channel and butterfly filter is:

$H_{\mathrm{total}}=H^{\′-1}H={\left(\mathrm{H\Φ}\right)}^{-1}\·H={\mathrm{\Φ}}^{-1}{\mathrm{HH}}^{-1}={\mathrm{\Φ}}^{-1}=\left[\begin{array}{cc}\mathrm{exo}(-{\mathrm{i\φ}}_x)& 0\\ 0& \exp \left({-\mathrm{i\φ}}_y\right)\end{array}\right]$

Notice that total transfer function successfully carried out polarization demultiplexing, just to x, the input signal on y polarization direction has been taken advantage of respectively a phase factor exp (i φ _x), exp (i φ _y), this phase factor can be conciliate in mode transfer piece 509 and remove at follow-up phase bit recovery.

Matrix inversion module 816 use H's ' is contrary, and weights initialize to butterfly filter.

The contrary of H ' is:

$H^{\′-1}={\left[\begin{array}{cc}{h}_1& h_2\\ h_3& h_4\end{array}\right]}^{-1}=\frac{1}{h_1{h}_4-h_2{h}_3}\left[\begin{array}{cc}{h}_4& -h_2\\ -h_3& h_1\end{array}\right]$

If the weight vector h of butterfly filter _xx, h _xy, h _yx, h _yylength is 1, establishes respectively initial value to be:

$h_{\mathrm{xx}}=\frac{h_4}{h_1{h}_4-h_2{h}_3},$ $h_{\mathrm{xy}}=\frac{-h_2}{h_1{h}_4-h_2{h}_3}$

$h_{\mathrm{yx}}=\frac{-h_3}{h_1{h}_4-h_2{h}_3},$ $h_{\mathrm{yy}}=\frac{h_1}{h_1{h}_4-h_2{h}_3}$

If, the weight vector h of butterfly filter _xx, h _xy, h _yx, h _yylength is greater than 1, final weight vector middle weights is made as to value above, and all the other locational weights are made as 0.The above is the frequency domain algorithm that carries out polarization demultiplexing with constant rotation sign training sequence.

According to the frequency spectrum R that receives signal _x(ω) 805 and R _y(ω) position 809,810 that in, peak value occurs, can estimate the frequency difference 811 of local oscillator (LO) 405 and signal carrier easily, and its method is as follows:

If: r is that reception signal angular frequency is ω _r; L is local laser output optical signal, and its baseband representation is:

Wherein ω _lfor the angular frequency of LO, ω _rsignal angular frequency, comprise LO and received the initial phase difference of signal and the phase noise of LO.

R and L are input to 90 ° of frequency mixers, and the signal after PD receives is:

Wherein, Δ ω=ω _l-ω _rit is the difference on the frequency of LO and signal.

As seen from the above analysis, in the time that the carrier frequency of LO and signal has difference on the frequency, receive signal can be on planisphere with the angular frequency rotation of-Δ ω.From frequency domain, the Frequency spectrum ratio that frequency difference makes PD receive signal receives frequency spectrum shift-Δ ω of signal r.If r is constant rotation sign training sequence, after DFT, its peak value will not appear at ω so _xor ω _yupper, but skew-Δ ω T _s, wherein T _sit is a symbol period.R _x(ω) at ω=ω _xnear peak is ω=ω ₁, should meet-Δ ω T _s=ω ₁-ω _x, can estimate frequency difference Δ f with following formula:

$\mathrm{\Δf}=-\frac{{\mathrm{\ω}}_1-{\mathrm{\ω}}_x}{2\mathrm{\π}{T}_s}$

Wherein T _sthe is-symbol cycle.Can also combine and use ω ₁, ω ₂, ω ₃, ω ₄improve the accuracy of frequency difference estimation.

Carry out the Time-Domain algorithm of polarization demultiplexing with constant rotation sign training sequence:

This algorithm belongs to least mean-square error (LMS) algorithm.As shown in Figure 9, LMS algoritic module 904 utilizes the input x of butterfly filter 903 to algorithm structure figure _p901, y _p902 and output x _out905, y _out906 upgrade the weights of butterfly filter 903.Iteration repeatedly after, the weights h of butterfly filter 903 _xx907, h _xy908, h _yx909, h _yy910 will converge on channel transfer functions inverse of a matrix, thereby successfully depolarization is multiplexing.The basic thought of Time-Domain algorithm is the character of utilizing constant rotation sign training sequence: in training sequence, the symbol phase sending on X polarization direction, rear one always than previous increase ω _x.The symbol phase sending on Y polarization direction, rear one than previous increase ω _y.In addition, symbol should be on the unit circle on planisphere.Therefore, the error signal of LMS algorithm is:

Wherein, x _outthe 905th, the output signal of current time butterfly filter x branch road, x _out' be a upper moment output signal, the output x of a upper defeated x branch road of moment _out' angle, ε _xfor the error of x branch road; y _outthe 906th, the output signal of current time butterfly filter y branch road, y _out' be a upper moment output signal, the output y of a upper defeated y branch road of moment _out' angle, ε _yfor the error of y branch road.

Upgrade the weights h of butterfly filter 903 with following formula _xx907, h _xy908, h _yx909, h _yy910:

$h_{\mathrm{xx}}\→h_{\mathrm{xx}}-\frac{\mathrm{\μ}}{2}\frac{d{\left|{\mathrm{\ϵ}}_x\right|}^2}{d{h}_{\mathrm{xx}}}=h_{\mathrm{xx}}+\mathrm{\μ}{\mathrm{\ϵ}}_x{x_p}^{*}$

$h_{\mathrm{xy}}\→h_{\mathrm{xy}}-\frac{\mathrm{\μ}}{2}\frac{d{\left|{\mathrm{\ϵ}}_x\right|}^2}{d{h}_{\mathrm{xy}}}=h_{\mathrm{xy}}+\mathrm{\μ}{\mathrm{\ϵ}}_x{x_p}^{*}$

$h_{\mathrm{yx}}\→h_{\mathrm{yx}}-\frac{\mathrm{\μ}}{2}\frac{d{\left|{\mathrm{\ϵ}}_y\right|}^2}{d{h}_{\mathrm{yx}}}=h_{\mathrm{yx}}+\mathrm{\μ}{\mathrm{\ϵ}}_y{x_p}^{*}$

$h_{\mathrm{yy}}\→h_{\mathrm{yy}}-\frac{\mathrm{\μ}}{2}\frac{d{\left|{\mathrm{\ϵ}}_y\right|}^2}{d{h}_{\mathrm{yy}}}=h_{\mathrm{yy}}+\mathrm{\μ}{\mathrm{\ϵ}}_y{y_p}^{*}$

Wherein, μ is the step-length of LMS algorithm, and * gets complex conjugate operation.

Figure 10 upgrades h _xx907 algorithm structure figure, upgrades h _xy908, h _yx909, h _yy910 algorithm similarly.The input x of butterfly filter _p1002 and output x _outthe 1001st, the input of update algorithm, ω _x1007 and μ 1012 are parameters of update algorithm.X _out1001 enter and get protractor 1003, and gained angle is through delayer 1006, then by adder 1008 and ω _x1007 are added, and addition result enters module 1009 and calculates x _out1001 enter opposite sign device 1005, then pass through adder 1010 with addition obtains x _p1002 enter conjugator 1004, then enter multiplier 1011 and parameter μ 1012 and ε _x1015 multiply each other obtains h _xxcorrection μ ε _xy _p ^*1016.Correction 1016 and h _xx1014 are added 1013, and addition result 1017 is as h _xxnew value.H _xxnew value output 1018 upgrade butterfly filter weights.The above is the Time-Domain algorithm of carrying out polarization demultiplexing with constant rotation sign training sequence.

Major technique advantage

The present invention is directed to the problem of polarization demultiplexing in the multiplexing coherent optical communication system of polarization diversity, designed a kind of special training sequence and be called and use constant rotation sign training sequence and related algorithm to comprise frequency domain algorithm and Time-Domain algorithm.Transport coherent receiver in this way and can carry out fast and automatically, exactly polarization demultiplexing.Use the most important advantage of constant rotation sign training sequence be receiving terminal do not need and the training sequence that sends synchronous, that is receiving terminal does not need to know the initial phase of transmission training sequence.

The present invention proposes two kinds of Processing Algorithm for constant rotation sign training sequence, is respectively frequency domain algorithm and Time-Domain algorithm.The advantage of frequency domain algorithm is directly to estimate according to the frequency spectrum that receives signal the weights of butterfly filter.If coherent receiver, in frequency domain compensation dispersion, must need to obtain receiving the frequency spectrum of signal so.Frequency domain algorithm can directly use this frequency spectrum, and does not need to increase the module of extra calculating frequency spectrum.

The advantage of Time-Domain algorithm is to use least mean-square error (LMS) algorithm to upgrade butterfly filter weights, and this algorithm complex is low.

Claims (2)

1. by the method for constant rotation sign training sequence polarization demultiplexing in the multiplexing coherent optical communication system of polarization diversity, it is characterized in that, said method comprising the steps of:

(1), at communication system initial phase, transmitting terminal sends respectively the constant rotation sign training sequence that velocity of rotation is different on two polarization directions;

(2) coherent receiver received training sequence it is processed, the weights of initialization butterfly filter, thus carry out polarization demultiplexing,

The method of the weights of described coherent receiver initialization butterfly filter has frequency domain algorithm and Time-Domain algorithm,

Frequency domain algorithm comprises following steps:

(1) ask discrete Fourier transform to calculate frequency spectrum to the training sequence receiving;

(2) from frequency spectrum, find peak value;

(3) go out channel transfer functions from peak estimation;

(4) with transfer function contrary come the weights of initialization butterfly filter,

Time-Domain algorithm comprises following steps:

(1) angle of upper moment butterfly filter two branch output signals of calculating;

(2) extrapolate the desired value of current time output signal from the angle of a upper moment output signal and the training sequence symbols velocity of rotation of polarization branch;

(3) difference of the actual value of the desired value of current time output signal and current time output signal is as the error signal of Time-Domain algorithm.

2. method according to claim 1, it is characterized in that, described training sequence is through rotating with constant speed along unit circle at its track of planisphere after symbol power normalization, and described velocity of rotation difference comprises that slewing rate is not identical or rotation direction is different.