CN101227266B - Method and apparatus for clock recovery - Google Patents

Abstract

The invention discloses a method and a device for recovering clocks, wherein the method comprises: receiving baseband symbols, adopting the fixed sampling rate to sample the baseband symbols, obtaining the baseband symbols, and adopting a B-spline fitting method to recover the baseband symbols on the position of a sampling point of a symbol clock for the obtained baseband symbols. The device comprises a receiver module which is used to receive the baseband symbols, a fixed adopting module, which is used to adopt the fixed sampling rate to obtain the baseband symbols and a re-sampling module which is used to adopt the B-spline fitting method to recover the baseband symbols on the position of the sampling point of the symbol clock for the obtained baseband symbols. The B-spline fitting method is not sensitive to frequency deviation of the carrier frequency since the B-spline fitting method is based on fitting calculation, and since noises damages to the sampling baseband symbols do not directly affect fitting of the B-spline, thereby the invention increases the ability of anti-frequency deviation and anti-interference, and the baseband symbols which are same with the sending end can be recovered more accurate .

Description

A kind of method of clock recovery and device

Technical field

The present invention relates to the technology of transmission data in the communications field, particularly a kind of method of all-digital receiver clock recovery and device.

Background technology

In digital communication system, in order to guarantee the correct transmission of data, must guarantee between the digital communication equipment synchronously, between the digital communication equipment comprise synchronously carrier frequency recovery synchronously and clock recovery synchronous.

At present, digital communication receiver is in order to recover baseband signalling accurately, and the transmitting terminal that needs and send this baseband signalling sends the corresponding to clock of frequency of baseband signalling, like this, could sample the baseband signal that obtains carrying to the baseband signalling that receives according to clock recovered.When digital communication receiver adopts the receiver of phase-locked loop, clock recovered is by an analogue device, and promptly voltage-controlled oscillator (VCO) produces, and receiver is by the frequency and the method for detecting phases of certain symbol clock, generation is to the control signal of VCO, thereby control VCO produces clock recovered.When digital communication receiver adopts all-digital receiver, in order to realize reliable performance and to produce in batches easily, eliminate the such analogue device of VCO, the method of clock recovery is all realized in a digit chip, with fixed sample rate the baseband signalling sampling that receives is obtained baseband signal, adopt method for resampling to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains again, thereby obtain the identical baseband signal that sends with transmitting terminal.

The principle of method for resampling is exactly from the signal that fixedly employing rate obtains, and through resampling, obtains the baseband signal at symbol clock sampling point position place.The typical embodiment of one of them of this method can be called the clock recovery method of Gardner, this method is utilized sampling time error detector (TED, Timing Error Detector) calculates corresponding interpolation point, the baseband signal of coming estimate symbol clock sampling point position in the Structure Calculation interpolation that utilize to be provided with.Usually, the method of sampling commonly used is an interpolation algorithm, Lagrange's interpolation algorithm for example, according to the Nyquist sampling principle, as long as sample frequency is greater than the bilateral bandwidth of received baseband signal, will contain sufficient signal message, thereby can recover the baseband signal that transmitting terminal sends undistortedly from sampled value.

Gardner clock recovery method based on Lagrange is a kind of method that adopts polynomial interopolation, it uses lagrange polynomial as interpolation polynomial, advantage is to calculate simply, can adopt the Farrow structure, shortcoming one is that the transition band width of low order interpolation filter compares broad, the speed of approaching is slow, and the calculating of high-order interpolation is too complicated, makes that like this low order interpolation frequency characteristic commonly used is bad; Shortcoming two is that Lagrange's interpolation imperial lattice phenomenon can occur when using the high order interpolation; Shortcoming three is Lagrange's interpolations when using n interpolation, the result that interpolation is come out with can be tending towards 0 along with the increase of n by the error of slotting function.

Since the shortcoming of the Garder clock recovery method of Lagrange, the present clock recovery method that also proposes a kind of B-spline interpolation, specific as follows described.

According to function theory, can approach of the linear combination of a continuous function with one group of function base, this group function base is called basic function so, and note is made p _{I, k}, wherein k represents the exponent number of this basic function, and i then is the location variable of basic function.Can approach the piecewise polynomial P that interpolation needs with the linear combination of this basic function _n(x), as formula (1)

$P_n\left(x\right)=\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{m}_i{p}_{i,k}\left(x\right),$ Formula (1)

M wherein _iBe constant, represent the linear combination weight of piecewise polynomial here.Obviously, formula (2) also need satisfy the condition of interpolation and is $P_n\left(x_j\right)=\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{m}_i{p}_{i,k}\left(x_j\right)=f_j,0\≤j\≤N,$ Formula (3)

Consider matrix form, make M=[m ₀..., m _N], F=[f ₀..., f _N],

In fact PM=F then, formula (4) interpolation problem here convert in order to seek basic function, and computing after this is unified for finding the solution the process of control point set M, be exactly to find the solution a system of linear equations in fact, interpolating function existence and unique sufficient and necessary condition are matrixes certainly

Nonsingular.

What usually adopt in the clock recovery process that resamples is 3 rank B-battens.

Consider with the B-batten to be that base is done function interpolation calculating at present, be without loss of generality, suppose that the node set that obtains is { nh} _{N ∈ Z}, the sigtnal interval is h=x _I+1-x _i, k represents the exponent number of B-spline base function, used basic function is

Consideration is at N+1 interpolation point { x ₀..., x _N}={ 0, h ..., the situation of Nh} is with N+1 function $P_{n,k}=B_k\left(\frac{x-\mathrm{nh}}{h}\right),n=\mathrm{0,1}...N$ Linear combination $P_k\left(x\right)=\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{m}_i{B}_k\left(\frac{x-\mathrm{ih}}{h}\right)$ Make interpolation, suppose that by inserting functional value be F=[f ₀..., f _N], according to B-spline function character and formula (4), matrix Become

Formula (5)

Symmetric property B _k(-x)=B _k(x) guarantee that P is a symmetrical matrix, by $B_k\left(x\right)=\underset{j=0}{\overset{k+1}{\mathrm{\Σ}}}\frac{{(-1)}^j}{k!}\left(\begin{array}{c}k+1\\ j\end{array}\right){(x+\frac{k+1}{2}-j)}_{+}^k$ Perhaps utilize de Boor-Cox recurrence formula, can calculate { B _k(n) } ₀ ^NValue, the most frequently used { B _k(n) } ₁ ⁴Numerical result as follows: $\left\{B_k\left(n\right)\right\}=\left\{\begin{array}{c}\left\{\mathrm{0,1,0}\right\},k=1\\ \left\{\mathrm{0.125,0.75,0.125}\right\},k=2\\ \left\{\mathrm{0,0.167,0.667,0.167,0}\right\},k=3\\ \left\{\mathrm{0.0026,0.1979,0.599,0.1979,0.0026}\right\},k=4,\end{array}\right.$ Formula (6)

By these values, just can produce matrix P soon and carry out interpolation calculation, obtain the interpolation of each interpolation point.

Special here, utilize

Carrying out interpolation is piecewise linear interpolation.

The process that adopts method for resampling to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains also needs to produce TED and filtering method, utilizes TED estimation and follow-up filtering method to calculate corresponding interpolation point.

Wherein, produce TED and can adopt Gardner at binary phase-shift keying/orthogonal phase shift keying (QPSK/BPSK) algorithm, under twice (with respect to the baseband signalling rate) over-sampling, its utilization $u_t\left(r\right)=y_I(r-\frac{1}{2})[y_I\left(r\right)-y_I(r-1)]+y_Q(r-\frac{1}{2})[y_Q\left(r\right)-y_Q(r-1)],$ Wherein, y _IBe the baseband signal of input, y _QBe the baseband signal that the employing fixed sample rate sampling of input obtains, r is a resample points, and as the estimation to the sampling clock deviation of BPS/QPSK, this estimator has very good performance.For the situation that does not have clock jitter, if | y _I(r)-y _I(r-1) | ≠ 0, then necessary $y_I(r-\frac{1}{2})=0,$ Equally, if | y _Q(r)-y _Q(r-1) | ≠ 0, then necessary $y_Q(r-\frac{1}{2})=0;$ This is the necessary condition of this estimation formulas as TED, satisfies for BPSK/QPSK.This method has simple and effective characteristics.

The method of B-spline interpolation is widely used in the middle of the digital communication system clock recovery resampling technology, but, because in digital communication welding system receiver, also comprise carrier frequency synchronization synchronously, the influence of general accurate estimation and counteracting carrier frequency offset all is after clock recovery, this means, clock recovery need be worked containing under the condition of certain frequency deviation, and the performance of the anti-frequency deviation of counterweight sampling process has proposed higher requirement like this; Have bursty interference in addition in digital communication system, interference signal may directly have been destroyed sampled signal, and at the enterprising row interpolation of the baseband signal of this sampling, diffusion can lead to errors.

In addition, at present the TED method of Gardner commonly used is to be derived by QPSK, for general Ditital modulation method, and QAM for example, this method can not directly be used.

Summary of the invention

The embodiment of the invention provides a kind of method of clock recovery in all-digital receiver, and this method can carry out improving in the process of clock recovery anti-frequency deviation and anti-jamming capacity, and more accurate recovering with transmitting terminal sends identical baseband signal.

The embodiment of the invention also provides a kind of device of clock recovery in all-digital receiver, and this device carries out improving in the process of clock recovery anti-frequency deviation and anti-jamming capacity, and more accurate recovering with transmitting terminal sends identical baseband signal.

According to above-mentioned purpose, technical scheme of the present invention is achieved in that

A kind of method of clock recovery, this method comprises: receive baseband signalling; Adopt fixed sample rate that described baseband signalling is sampled, obtain baseband signal; Adopt B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains; The process that described employing B-spline-fit method recovers the baseband signal at symbol clock sampling point position place comprises: adopt the value of the baseband signal that three rank B-spline-fit methods calculate, the value of the described baseband signal that calculates is inserted into corresponding symbol clock sampling point position; The value process of the described baseband signal that calculates comprises: calculate C=BF, wherein F is the discrete function on the discrete point of the described baseband signal that obtains, B is that three rank spline functions are at the different time shift value, described different time shift value linear independence, C is first intermediate variable of the value of calculating baseband signal; Calculate A=Φ ^-1C, wherein Φ is that selected exponent number is the matrix of n*n, and n is a natural number, obtains A, and A is second intermediate variable of the value of calculating baseband signal; 4 respectively corresponding 4 battens in the middle of from A, selecting

Coefficient, with [the x that obtains _i, x _I+1] interval value is as the value of the described baseband signal that calculates.

A kind of device of clock recovery: receiver module, fixed sample module and resampling module, wherein, receiver module is used to receive baseband signalling; The fixed sample module is used to adopt fixed sample rate to sample and obtains baseband signal; The resampling module is used for adopting B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains; Described resampling module comprises fitting module and interpolating module, and wherein, fitting module is used to adopt the value of the baseband signal that three rank B-spline-fit methods calculate; Interpolating module, be used for obtaining symbol clock sampling point position place according to sampling time error detector TED method, the value of the described baseband signal that calculates is inserted into corresponding symbol clock sampling point position, recovers the baseband signal at symbol clock sampling point position place; The value process of the described baseband signal that calculates comprises: calculate C=BF, wherein F is the discrete function on the discrete point of the described baseband signal that obtains, B is that three rank spline functions are at the different time shift value, described different time shift value linear independence, C is first intermediate variable of the value of calculating baseband signal; Calculate A=Φ ^-1C, wherein Φ is that selected exponent number is the matrix of n*n, and n is a natural number, obtains A, and A is second intermediate variable of the value of calculating baseband signal; 4 respectively corresponding 4 battens in the middle of from A, selecting

Coefficient, with [the x that obtains _i, x _I+1] interval value is as the value of the described baseband signal that calculates.

From such scheme as can be seen, method that the embodiment of the invention provides and device are in the process of the baseband signal that recovers symbol clock sampling point position place, adopt the method for B-spline-fit, the method of B-spline-fit is because based on The Fitting Calculation, frequency deviation to carrier frequency is insensitive, and because noise can directly not influence the match of B-batten to the destruction of the baseband signal of sampling.Therefore, method that the embodiment of the invention provides and device improve anti-frequency deviation and anti-jamming capacity in the process of carrying out clock recovery, and more accurate recovering with transmitting terminal sends identical baseband signal.

Description of drawings

The method flow diagram of clock recovery in the all-digital receiver that Fig. 1 provides for the embodiment of the invention;

Fig. 2 is the result schematic diagram that the different time displacement of the embodiment of the invention three rank B-spline functions obtains;

The fast method flow chart of the employing three rank B-spline-fits that Fig. 3 provides for the invention process;

The apparatus structure schematic diagram of clock recovery in the all-digital receiver that Fig. 4 provides for the embodiment of the invention.

Embodiment

In order to make the purpose, technical solutions and advantages of the present invention clearer, below lift specific embodiment and, the present invention is described in more detail with reference to accompanying drawing.

In all-digital receiver, the process that adopts method for resampling to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains comprises: generate one with the consistent local clock of transmitting terminal baseband signal rate, this local clock is generated by digital controlled oscillator (NCO), the control signal of control NCO is produced by TED, the clock that has had NCO to produce just can adopt method for resampling to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains.Therefore, in the process of carrying out clock recovery in order to improve anti-frequency deviation and anti-jamming capacity, more accurate recovering with transmitting terminal sends identical baseband signal, the embodiment of the invention adopts B-spline-fit method the baseband signal that obtains to be recovered the baseband signal at symbol clock sampling point position place in the process to the baseband signal that recovers symbol clock sampling point position place.

The embodiment of the invention also will have the TED method in the technology to expand in clock recovery TED method, obtain being applicable to the TED method of any modulation.

The method flow diagram of clock recovery in the all-digital receiver that Fig. 1 provides for the embodiment of the invention, its concrete steps comprise:

Step 101, receive baseband signalling after, adopt fixed sample rate to sample and obtain baseband signal;

Step 102, adopt B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains.

In this step, adopt three rank B-spline-fit methods usually, this B-spline-fit method can adopt the solution procedure of carrying out fitting coefficient based on FIR filtering, calculates simply numerical stability.

In this step, adopt B-spline-fit method to recover the baseband signal at symbol clock sampling point position place, and B-spline-fit method is because based on The Fitting Calculation, frequency deviation to carrier frequency is insensitive, and, send identical baseband signal so can accurately recover with transmitting terminal because noise can directly not influence the match of B-batten to the destruction of sampled point.

In this step, adopt three rank B-spline-fit methods to recover the baseband signal at symbol clock sampling point position place, in fact be exactly to adopt three rank B-spline-fit methods to calculate a value, described value is inserted into symbol clock sampling point position place, wherein, corresponding interpolation point can adopt the TED of prior art to produce, and the TED that also can adopt the embodiment of the invention to provide produces.

Below lift the step 102 that the detailed explanation embodiment of the invention of specific embodiment provides, in this embodiment, adopt the B-batten to carry out the method for least square fitting.

At first, introduce the least square ratio juris.If f is for obtaining discrete function given on the discrete point of baseband signal, (x at m+1 _k, f (x _k)), k=0,1 ... m, least square method makes for asking s: $\underset{j=0}{\overset{m}{\mathrm{\Σ}}}{[f\left(x_j\right)-s\left(x_j\right)]}^2$ Get minimum.S is the least square solution of f on the discrete point of m+1 baseband signal.If

₀, ₁, _n, be the linear independence group of functions, belong to the multinomial of exponent number smaller or equal to n, s is expressed as:

Then the problem of this embodiment is asked a exactly _i, i=0,1 ... n makes

Minimum, as can be seen, it is a _i, i=0,1 ... the quadratic function of n, utilize the function of many variables to ask the extreme value necessary condition to have: $\frac{\∂F}{{\∂\mathrm{\α}}_k}=0,k=\mathrm{0,1},...,n,$ Promptly

K=0,1 ... so n

K=0,1 ... n, because

₀,

₁,

_n, so linear independence is a _i, i=0,1 ... the coefficient matrix of n is nonsingular, and unique solution is arranged.

This embodiment uses three rank B-battens to carry out match, and the combination of function of three rank B-battens is adopted in expectation

₀,

₁,

_nFunction of match on the discrete point of the baseband signal that m+1 obtains is expressed as: Wherein

₀,

₁, _nBe the result that the different time displacement of three rank B-spline functions obtains, as shown in Figure 2.

The formula of utilization least square method:

K=0,1 ... n can write matrix form: Φ A=BF, wherein F=[f (x ₀) f (x ₁) f (x ₂) ... f (x _m)] ^T,

Wherein

For example, m=10, n=7, during with B-three rank battens:

$B=\left[\begin{array}{cccccccccccc}0& 1/6& 2/3& 1/6& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 1/6& 2/3& 1/6& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 1/6& 2/3& 1/6& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 1/6& 2/3& 1/6& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 1/6& 2/3& 1/6& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 1/6& 2/3& 1/6& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 1/6& 2/3& 1/6& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 1/6& 2/3& 1/6& 0\end{array}\right]$

$\mathrm{\Φ}=\left[\begin{array}{cccccccc}1/2& 2/9& 1/36& 0& 0& 0& 0& 0\\ 2/9& 1/2& 2/9& 1/36& 0& 0& 0& 0\\ 1/36& 2/9& 1/2& 2/9& 1/36& 0& 0& 0\\ 0& 1/36& 2/9& 1/2& 2/9& 1/36& 0& 0\\ 0& 0& 1/36& 2/9& 1/2& 2/9& 1/36& 0\\ 0& 0& 0& 1/36& 2/9& 1/2& 2/9& 1/36\\ 0& 0& 0& 0& 1/36& 2/9& 1/2& 2/9\\ 0& 0& 0& 0& 0& 1/36& 2/9& 1/2\end{array}\right]$

The group of solving an equation Φ A=BF, wherein f (x ₀) f (x _m) value reality do not use, just can obtain coefficient A=[a ₀a ₁a ₂A _n] ^T, again with the coefficient substitution

Just can obtain fitting result, though as can be seen from Figure 2, [x _i, x _I+1] 4 battens only need being correlated with of interval value

_I-3, _I-2,

_I-1, _iJust can obtain with their coefficient, if but the size of equation group Φ A=BF is not enough, and the coefficient precision of trying to achieve is not enough, and the match error of coming out is very big, because in fact B and Φ matrix should be that infinite dimension could realize accurate match.

Calculate the B-spline-fit and can adopt the method for pre-filtering to carry out, calculate simply, and accuracy guarantee.For example for the finding the solution of Φ A=BF in the process of calculating the B-spline-fit, because could realize the calculating of accurate fitting coefficient when in fact having only B and Φ matrix to be infinite dimension.In order to find the solution A=Φ ^-1BF, calculation procedure comprises as follows:

1), calculates C=BF, present embodiment adopts existing 3 rank FIR filters [1/6 2/3 1/6], input signal is carried out filtering, wherein F obtains discrete function given on the discrete point of baseband signal, B be three rank spline functions at the different time shift value, described different time shift value is the linear independence functional value;

2), find the solution A=Φ ^-1C, directly do not calculate inverse matrix, matrix Φ be actually an infinite dimension matrix get in the middle of the capable n+1 of the n+1 row of symmetry obtain, Φ matrix each provisional capital in the middle of in fact is the result of the displacement of vector [1,/36 2/9 1/2 2/9 1/36], has removed the element of head or tail except several row at edge.When matrix Φ was very big, its inverse matrix also had such characteristic, and middle provisional capital is the result of the difference displacement of same vector (being defined as L).Therefore, from main inverse of a matrix matrix to angle both sides logarithmic decrement also is main to angle both sides logarithmic decrement, Φ inverse of a matrix matrix is main to angle both sides logarithmic decrement like this, vectorial in other words L from the centre to the both sides logarithmic decrement, the data at the edge of L decay to 0 very soon, and it is useful having only limited several.So this embodiment adopts a fixed coefficient FIR filter of L correspondence to come the filtering to C,, ask the process of A very simple like this: in conjunction with the 1st along with the displacement of time obtains A) step, to the baseband signal F=[f (x that obtains ₀) f (x ₁) f (x ₂) ... f (x _m)] ^T,, just obtain A with a fixed coefficient FIR filter filtering setting vectorial L correspondence more earlier with 3 rank FIR filters [1/6 2/3 1/6] filtering.

3), from A, select 4 middle respectively corresponding 4 battens

_I-3,

_I-2,

_2i-1,

_iCoefficient, just can match [x _i, x _I+1] interval interpolation.

Because corresponding interpolation point, promptly respective symbol clock sampling point position can adopt the TED of prior art and follow-up filtering to generate, and also can adopt TED that the embodiment of the invention provides and follow-up filtering generation.Wherein, how carrying out filtering is the filtering method that can adopt prior art, is not repeated here.In embodiments of the present invention, TED is expanded, make it be adapted to any modulation TED method in the digital communication system.

In this embodiment, can utilize existing TED method: $u_t\left(r\right)=y_I(r-\frac{1}{2})[y_I\left(r\right)-y_I(r-1)]+y_Q(r-\frac{1}{2})[y_q\left(r\right)-y_Q(r-1)],$ Wherein, y _IBe the baseband signal of input, y _QBe the baseband signal through resampling of input, r is a resample points, and as the estimation to the sampling clock deviation of BPSK/QPSK, this estimator has very good performance.It is not difficult to find after careful observation for the situation that does not have clock jitter, if | y _I(r)-y _I(r-1) | ≠ 0, then necessary $y_I(r-\frac{1}{2})=0,$ Equally, if | y _Q(r)-y _Q(r-1) | ≠ 0, then necessary $y_Q(r-\frac{1}{2})=0;$ This is the necessary condition of this estimation formulas as the TED method, satisfies for BPSK/QPSK, but for the modulation in the general digital communication system, the TED method of Gardener can be utilized the center displacement to be promoted, and is revised as the method that this embodiment proposes:

$u_t^{\mathrm{universal}}\left(r\right)=\{y_I(r-\frac{1}{2})-\frac{y_I\left(r\right)+y_I(r-1)}{2}\}[y_I\left(r\right)-y_I(r-1)]$

$+\{y_Q(r-\frac{1}{2})-\frac{y_Q\left(r\right)+y_Q(r-1)}{2}\}[y_Q\left(r\right)-y_Q(r-1)],$

Verify its validity for BPSK/QPSK, if | y _I(r)-y _I(r-1) | ≠ 0, then necessarily have $\frac{y_I\left(r\right)+y_I(r-1)}{2}=0,$ And this moment $y_I(r-\frac{1}{2})=0,$ Certainly condition is not have clock jitter; Equally, if | y _Q(r)-y _Q(r-1) | ≠ 0, then $\frac{y_Q\left(r\right)+y_Q(r-1)}{2}=0,$ And have this moment equally $y_Q(r-\frac{1}{2})=0,$ When having clock jitter, do not meet the requirements fully; And, can verify that this estimator satisfies the necessary condition as no inclined to one side TED method for the modulation in the general digital communication system.

The fast method flow chart of the employing three rank B-spline-fits that Fig. 3 provides for the invention process, the process that adopts three rank B-spline-fits is exactly the process of solving equation group Φ A=BF, wherein F obtains discrete function given on the discrete point of baseband signal, B be three rank spline functions at the different time shift value, its concrete steps comprise:

Step 301, calculating C=BF use 3 rank FIR filters [1/6 2/3 1/6] here, and F does filtering to input signal;

Step 302, exponent number of selection are n*n, and fully big matrix Φ calculates Φ in advance ^-1, get the in the middle of it

N/2

OK,

Represent rounding operation, get a middle K coefficient of this journey,, use this FIR filter that the result that step 501 obtains is carried out filtering, just can obtain the coefficient A that is asked as the coefficient of fixed coefficient FIR filter;

In this embodiment, the quantity of K depends on required precision, and K is big more, and precision is high more, and the more little precision of K is poor more, can be selected by test;

Step 303,4 respectively corresponding 4 battens in the middle of from A, selecting

_I-3,

_I-2,

_I-1,

_iCoefficient, obtain [x _i, x _I+1] interval interpolation, with this [x that obtains _i, x _I+1] interval value insertion [x _i, x _I+1] interval, recover the baseband signal at symbol clock sampling point position place.

The embodiment of the invention also provides the device of clock recovery in a kind of all-digital receiver, as shown in Figure 4, the apparatus structure schematic diagram of clock recovery comprises: receiver module, fixed sample module and resampling module in the all-digital receiver that Fig. 4 provides for the embodiment of the invention, wherein

Receiver module is used for receiving baseband signal;

The fixing module that adopts is used to adopt fixed sample rate to sample and obtains baseband signal;

The resampling module is used for adopting B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains.

In this embodiment, the resampling module comprises fitting module and interpolating module, wherein,

Fitting module is used to adopt three rank B-spline-fit methods to obtain the value at corresponding symbol clock sampling point position place;

Interpolating module is used for obtaining symbol clock sampling point position place according to the TED method, and the value of described correspondence is inserted into symbol clock sampling point position place, recovers the baseband signal at symbol clock sampling point position place.

Method that the embodiment of the invention provides and device use the method for B-spline-fit in the clock recovery of carrying out all-digital demodulation resamples, have the advantage of match itself, and be insensitive to frequency deviation and burst noise.Adopt the algorithm of asking fitting coefficient based on fixed coefficient FIR filtering method in the method for B-spline-fit, have good stability, the numerical precision controllability is good, and is simple in structure, the advantage that amount of calculation is little.The coefficient number K that this embodiment selects when FIR filtering, K is big more, and numerical precision is high more, and the more little calculating of K is simple more, selects flexibly according to the needs of numerical precision, and in computational process, all adopt the FIR filter structure, guarantee linear-phase filtering and numerical stability.

In addition, the embodiment of the invention has just obtained going for the TED algorithm for estimating of any modulation with the method for the TED method utilization of the Gardner of prior art simple center displacement.

The above only is preferred embodiment of the present invention, not in order to restriction the present invention, all any modifications of being made within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. the method for a clock recovery is characterized in that, this method comprises:

Receive baseband signalling;

Adopt fixed sample rate that described baseband signalling is sampled, obtain baseband signal;

Adopt B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains;

The process that described employing B-spline-fit method recovers the baseband signal at symbol clock sampling point position place comprises:

Adopt the value of the baseband signal that three rank B-spline-fit methods calculate, the value of the described baseband signal that calculates is inserted into corresponding symbol clock sampling point position;

The value process of the described baseband signal that calculates comprises: calculate C=BF, wherein F is the discrete function on the discrete point of the described baseband signal that obtains, B is that three rank spline functions are at the different time shift value, described different time shift value linear independence, C is first intermediate variable of the value of calculating baseband signal;

Calculate A=Φ ^-1C, wherein Φ is that selected exponent number is the matrix of n*n, and n is a natural number, obtains A, and A is second intermediate variable of the value of calculating baseband signal;

4 respectively corresponding 4 battens in the middle of from A, selecting

Coefficient, with [the x that obtains _i, x _I+1] interval value is as the value of the described baseband signal that calculates.

2. the method for claim 1 is characterized in that, the matrix that described matrix Φ decays for the therefrom mind-set both sides that the employing least square method calculates.

3. the method for claim 1 is characterized in that, described calculating C=BF adopts three rank FIR filtering to carry out.

4. the method for claim 1 is characterized in that, described calculating A=Φ ^-1The process of C comprises:

Calculate Φ ^-1, get the in the middle of it

OK, The expression rounding operation is chosen the middle K coefficient of this journey according to the scope of setting, and according to this coefficient use FIR filter the C that obtains is carried out filtering.

5. the method for claim 1 is characterized in that, described symbol clock sampling point position place adopts sampling time error detector method and obtains through after the filtering, wherein, adopts the process of sampling time error detector method to comprise:

$u_t^{\mathrm{universal}}\left(r\right)=\{y_I(r-\frac{1}{2})-\frac{y_I\left(r\right)+y_I(r-1)}{2}\}[y_I\left(r\right)-y_I(r-1)]$ $+\{y_Q(r-\frac{1}{2})-\frac{y_Q\left(r\right)+y_Q(r-1)}{2}\}[y_Q\left(r\right)-y_Q(r-1)],$ Wherein, y _IBe the baseband signal of input, y _QBe the baseband signal through resampling of input, r is a resample points.

6. the device of a clock recovery is characterized in that, this device comprises: receiver module, fixed sample module and resampling module, wherein,

Receiver module is used to receive baseband signalling;

The fixed sample module is used to adopt fixed sample rate to sample and obtains baseband signal;

The resampling module is used for adopting B-spline-fit method to recover the baseband signal at symbol clock sampling point position place to the baseband signal that obtains;

Described resampling module comprises fitting module and interpolating module, wherein,

Fitting module is used to adopt the value of the baseband signal that three rank B-spline-fit methods calculate;

Interpolating module, be used for obtaining symbol clock sampling point position place according to sampling time error detector TED method, the value of the described baseband signal that calculates is inserted into corresponding symbol clock sampling point position, recovers the baseband signal at symbol clock sampling point position place;

Described fitting module specifically is used for adopting according to following step the value of the baseband signal that three rank B-spline-fit methods calculate: calculate C=BF, wherein F is the discrete function on the discrete point of the described baseband signal that obtains, B is that three rank spline functions are at the different time shift value, described different time shift value linear independence, C is first intermediate variable of the value of calculating baseband signal; Calculate A=Φ ^-1C, wherein Φ is that selected exponent number is the matrix of n*n, and n is a natural number, obtains A, and A is second intermediate variable of the value of calculating baseband signal; 4 respectively corresponding 4 battens in the middle of from A, selecting

Coefficient, with [the x that obtains _i, x _I+1] interval value is as the value of the described baseband signal that calculates.

7. device as claimed in claim 6 is characterized in that, described symbol clock sampling point position place adopts sampling time error detector method and obtains through after the filtering, and described interpolating module is used for carrying out the sampling time error detection according to following formula:

$u_t^{\mathrm{universal}}\left(r\right)=\{y_I(r-\frac{1}{2})-\frac{y_I\left(r\right)+y_I(r-1)}{2}\}[y_I\left(r\right)-y_I(r-1)]$ $+\{y_Q(r-\frac{1}{2})-\frac{y_Q\left(r\right)+y_Q(r-1)}{2}\}[y_Q\left(r\right)-y_Q(r-1)],$ Wherein, y _IBe the baseband signal of input, y _QBe the baseband signal through resampling of input, r is a resample points.

8. device as claimed in claim 6 is characterized in that, the matrix that described matrix Φ decays for the therefrom mind-set both sides that the employing least square method calculates.

9. device as claimed in claim 6 is characterized in that, described calculating C=BF adopts three rank FIR filtering to carry out.

10. device as claimed in claim 6 is characterized in that, described fitting module is calculated A=Φ ^-1The process of C comprises:

Calculate Φ ^-1, get the in the middle of it

OK,

The expression rounding operation is chosen the middle K coefficient of this journey according to the scope of setting, and according to this coefficient use FIR filter the C that obtains is carried out filtering.

Images