CN1599367A - Synchronous method of orthogonal freuency division multiplex in broadband radio insertion system - Google Patents

Abstract

A synchronization method in OFDM broadband wireless access system includes such steps as utilizing NO.1 frame head as a delay related decision to realize rough synchronization, using the conjugate symmetry of No.2 frame head to realize accurate timing, the frame heads roughly and accurately estimate the decimal frequency deviation separately, the frequency field sequences of the two heads are used to estimate the integer frequency bias and finally synchronous tracing.

Description

The method for synchronous of OFDM in the BWA

Technical field

The present invention relates to a kind of synchronous implementation of OFDM (OFDM) of the 802.16a of being used in particular for broadband radio access network, belong to communication technical field.

Background technology

Antijamming capability is strong, the frequency efficiency advantages of higher more and more obtains extensive concern because of it has ofdm system, and wireless broadband access system 806.16a has just adopted the OFDM technology as one of transmission means of system.But it requires very high to sign synchronization: OFDM uses the subcarrier of quadrature to improve the availability of frequency spectrum, if the regularly asynchronous orthogonality that can destroy subcarrier; Frequency shift (FS) causes the phase place rotation and the amplitude fading of useful signal, more seriously causes between subcarrier and disturbs, and systematic function will descend significantly, thereby compensation is because the asynchronous systematic function of bringing with frequency of time descends very important.

In recent years, people have carried out extensive studies synchronously to OFDM, wherein more typical algorithm mainly comprises following: Moose has provided the frequency domain maximum Likelihood and has estimated frequency deviation, its essence is and utilize the frequency domain that repeats of symbol to estimate, Moose is this to provide basic thinking based on the thought that repeats to be correlated with for afterwards frequency offset estimating algorithm, but because his algorithm requires frame head to use the symbol of two repetitions and channel coherence time at least greater than duration of two symbols, so now much in the real systems and without the method; Nogami and Nagashima have adopted a kind of null symbol, descend and come frame synchronization by detecting power at receiving terminal, but this method is applicable to continuous structure, for the bursty data structure, should not adopt the method owing to can't correctly distinguishing Null symbol and free time; Van de Beek has proposed the maximal possibility estimation algorithm based on CP, this algorithm is unable to estimate integer frequency bias, and can not provide frame synchronization information, thereby be used for regularly and the tracking phase of frequency deviation more, consider that synchronous tracking will guarantee that algorithm is simple, have many documents all to propose various improvement algorithms on this basis; M.Schmidl has proposed the method based on training sequence, and he utilizes two training sequences of frame head to realize energy measuring and synchronously smart and frequency offset estimating, the fairly simple and easy realization of this algorithm; Classen has also proposed based on pilot frequency system, and its prerequisite is slow time varying channel, and this algorithm has been introduced ICI when frequency domain is estimated frequency deviation, so performance descends to some extent; Also have the blind estimation of studying at present also more, because its average based on a plurality of symbols become when requiring channel slow, and the algorithm complexity is used for frequency offset tracking more.

Summary of the invention

Technical problem: the method for synchronous that the purpose of this invention is to provide OFDM in a kind of BWA, this method in system design based on training sequence, make full use of the information that frame head provides, to essence regularly synchronously and the integer frequency bias method of estimation improve, adjusted synchronizing sequence, thereby reduced amount of calculation, and improved the frequency offset estimating performance.

Technical scheme: the method for synchronous in the ofdm system of the present invention: utilize frame head one to do and postpone relevant judgement, realize synchronously thick; That utilizes frame head two grips symmetry altogether, realizes regularly smart; Carry out the decimal frequency bias rough estimate with frame head one, carry out the smart estimation of decimal frequency bias with frame head two; Estimate integer frequency bias with the frame head two and the frequency domain sequence of frame head one; Last tracking synchronously.

Wherein: frame head has comprised long training sequence and short training sequence, frame head 1 comprises the cycle of 4 repetitions for short training sequence, and frame head 2 comprises 2 repetition periods for long training sequence, and this repetition period property generates by inserted 0 before anti-fast fourier transform IFFT conversion.

Thick synchronous method is: the periodicity of utilizing frame head one, adopt and postpone correlation technique (promptly asking a signal and its correlation function) through signal after the time delay, two sliding windows are arranged, one is the correlation function that calculates received signal and its inhibit signal, and retardation coefficient equals the Cycle Length 64 of Short Training symbol; The signal energy that another calculates in associated window is to be used for judgment variables is carried out normalized, and judgment variables just is no longer dependent on the abswolute level of energy like this, when judgment variables during greater than threshold value, this place can be defined as just locating synchronously.

Smart method regularly is: because frame head two of the present invention not only has cycle repeatability, and have and grip symmetry altogether, the frame head structure is as [A/4 B/4 A/4 B/4], and B grips symmetry altogether about A.Utilize this symmetry of gripping altogether just, that asks the first half frame head and half frame head of back grips correlation function altogether as judgment variables, searches for the maximum of this judgment variables, and this maximum place is smart place, the just accurate starting point of a frame synchronously.Smart timing algorithm is than asking cross correlation algorithm amount of calculation itself just to reduce the multiplication of half, and not because search integer frequency bias and repeated calculation, amount of calculation significantly reduces especially, the another one benefit is utilized exactly to grip altogether when symmetry is asked maximum correlation peak and not influenced by frequency deviation, has improved synchronous precision.

The decimal frequency bias estimation approach is: carry out the decimal frequency bias rough estimate with frame head one, carry out the smart estimation of decimal frequency bias with frame head two.For short training sequence, estimable maximum frequency deviation is 2; For long training sequence, its D=126, estimable maximum frequency deviation is 1.

The integer frequency bias estimation approach is: adopt frame head two and frame head one frequency-domain combined estimation integer frequency bias.Because ICI has been avoided in the correction of decimal frequency bias, thereby frequency domain symbol is only because integer frequency bias produces displacement, and frame head one is consistent with the displacement of frame head two, and mobile figure place is 4 multiple; This method utilizes the ratio on frame head two and the frame head one frequency domain 4 gall nut carrier waves to form new known PN sequence v _k, with the corresponding PN sequence that receives do relevant, 4 sampled points search of each displacement maximums in possible integer frequency bias scope, the carry digit when determining maximum is the integer frequency bias coefficient.

Synchronously the method for following the tracks of by: utilize the feature of information redundancy in the Cyclic Prefix intrinsic in the ofdm signal, adopt maximum likelihood estimator module time domain carry out symbol regularly and frequency deviation unite estimation.

Beneficial effect: the characteristics that the present invention is based on the 802.16a system have proposed the whole method for synchronous of a cover, not only reduced algorithm complex, and improved the performance of system, wherein regularly smart synchronized algorithm utilized the 802.16a frame head grip symmetry altogether, littler and insensitive than traditional smart timing method amount of calculation to frequency deviation, thereby can before Frequency Synchronization, finish, make variously to have overcome interdependency between synchronously, improved the robustness of system.This method based on training sequence, makes full use of the information that frame head provides in system design, to essence regularly synchronously and the integer frequency bias method of estimation improve, adjusted synchronizing sequence, thereby reduced amount of calculation, and improved the frequency offset estimating performance.

Description of drawings

Fig. 1 is a frame head structure chart of the present invention.

Fig. 2 is synchronization scenario figure of the present invention.

Fig. 3 is that the present invention postpones relevant schematic diagram.

Fig. 4 is the thick block diagram synchronously of the present invention.

Fig. 5 is that the present invention postpones relevant judgment variables response diagram.

Fig. 6 is the smart block diagram synchronously of the present invention.

Fig. 7 is a maximum likelihood estimator module structure chart of the present invention.

Fig. 8 is that the present invention follows the tracks of block diagram synchronously.

Have among the above figure: postpone relevant sliding window C, signal energy computation sliding window P, c (n) is receiving signal delayed correlation function, and r (n) is the signal that receives, and p (n) is the received signal energy function, and M (n) is the time synchronized function, Z ^-DBe retardation coefficient ,+for addition, * for multiplying each other, ÷ is for being divided by, || ²For asking absolute square, () ²For asking square () ^*Grip θ for asking altogether _MLBe symbol timing, ε _MLBe frequency offset estimating, γ (θ) is that received signal N position postpones correlation function, and Φ () is the result of displacement summation, ρ || ²/ 2 for asking half of absolute square, || for asking absolute value, ∠ is for asking angle, and-1/2 π is the multiple of signal times, and Arg max is a maximizing, and r (k) is the signal that receives.

Embodiment

Frame head structure of the present invention as shown in Figure 1, long training sequence and short training sequence have been comprised, wherein frame head 1 comprises the cycle of 4 repetitions for short training sequence, there are 64 sample points in each cycle, and frame head 2 comprises 2 repetition periods for long training sequence, there are 128 sample points in each cycle, and wherein CP is a Cyclic Prefix.

Concrete scheme such as Fig. 2 mainly are divided into following five steps:

(1) synchronously thick

The thick periodicity of having utilized frame head one synchronously, this algorithm can be described as the delay related algorithm, as shown in Figure 3.

Show two sliding window C and P among the figure, the C window is the correlation function of received signal and its inhibit signal, therefore is called to postpone to be correlated with, and retardation coefficient Z-D equals the Cycle Length 64 of Short Training symbol in the present invention; The P window carries out the signal energy computation in the associated window, is to be used for judgment variables is carried out normalized, and judgment variables just is no longer dependent on the abswolute level of energy like this.

If c (n) is receiving signal delayed correlation function, r (n) is the signal that receives, and p (n) is the received signal energy function, M (n) is the time synchronized function, L is the length of window, generally gets the length of each sequence in the frame head repetitive sequence, the rough estimate position n that Frame begins _Coarse

$c\left(n\right)=\underset{k=0}{\overset{L-1}{\mathrm{\Σ}}}r(n+k)r*(n+k+D)---\left(1\right)$

$p\left(n\right)=\underset{k=0}{\overset{L-1}{\mathrm{\Σ}}}{\left|r(n+k+D)\right|}^2---\left(2\right)$

The time synchronized function definition is

$M\left(n\right)=\frac{{\left|c\left(n\right)\right|}^2}{{\left(p\left(n\right)\right)}^2}---\left(3\right)$

Because c is for postponing relevant sliding window, p is a signal calculated energy sliding window, and the way of iteration can be used for reducing amount of calculation.

Specifically as shown in Figure 4: utilize preceding 64 data point calculation c (n) of the data r (n) that receives earlier, p (n) and M (n) value, promptly $c\left(1\right)=\underset{k=0}{\overset{63}{\mathrm{\Σ}}}r(1+k)r^{*}(1+k+64),$ $p\left(1\right)=\underset{k=0}{\overset{63}{\mathrm{\Σ}}}{\left|r(1+k+64)\right|}^2$ With

$M\left(1\right)=\frac{{\left|c\left(1\right)\right|}^2}{{\left(p\left(1\right)\right)}^2};$

Carry out iteration then:

c(2)＝c(1)-r(1)r ^*(1+64)+r(65)r ^*(65+64)

p(2)＝p(1)-|r(1+64)| ²+|r(65+64)| ²

$M\left(2\right)=\frac{{\left|c\left(2\right)\right|}^2}{{\left(p\left(2\right)\right)}^2}$

By that analogy, can calculate one group of (supposing 1000) M (n) value.Fig. 5 is targeting sequencing of the present invention postpones relevant judgment variables under the SNR=10dB environment response, and wherein transverse axis is n, and the longitudinal axis is M (n).When Frame began, M (n) jumped to maximum rapidly as seen from the figure, and this saltus step can very effectively be used for the rough estimate position n that the decision data frame begins _CoarseThis method is subjected to the statistical property and the noise effect of Short Training symbol sebolic addressing self, and irrelevant with frequency deviation, so carried out synchronously thick before Frequency Synchronization.Experimental results show that under the situation of SNR＞6dB, the thick net synchronization capability of system is more satisfactory.

(2) synchronously smart

Frame head two of the present invention not only has cycle repeatability, and has and grip symmetry altogether, and the frame head structure is as [A/4 B/4 A/4 B/4], and B grips symmetry altogether about A.Utilize this symmetry of gripping altogether just, smart timing algorithm is done following improvement:

If P (d) is the received signal auto-correlation function, R (d) is the received signal energy function, and r (n) is the signal that receives.N is the number of the data in the OFDM symbol, M _Pro(d) be the time synchronized function of smart synchronized algorithm.

$M_{\mathrm{Pro}}\left(d\right)=\frac{{\left|P\left(d\right)\right|}^2}{{\left(R\left(d\right)\right)}^2}---\left(4\right)$

Wherein, $P\left(d\right)=\underset{k=0}{\overset{N/4-1}{\mathrm{\Σ}}}r(d-k)r(d+k)---\left(5\right)$

$R\left(d\right)=\underset{k=0}{\overset{N/4-1}{\mathrm{\Σ}}}{\left|r(d+k)\right|}^2---\left(6\right)$

This moment, smart timing algorithm was than asking cross correlation algorithm amount of calculation itself just to reduce the multiplication of half, and the repeated calculation because search for integer frequency bias not, and amount of calculation significantly reduces especially.

Concrete steps are as shown in Figure 6: according to the rough estimate sync bit n that slightly obtains synchronously _Coarse, the position that obtains first bit data behind frame head two CP is (n _Coarse+ 32+256+32)=n _Coarse+ 320, and frame head two is to be n with the 65th bit data _Coarse+ 320+64=n _Coarse+ 384 places are the center symmetry, thick also mention the position of estimating inclined to one side morning in synchronously, so calculate $R(n_{\mathrm{coarse}}+384)=\underset{k=0}{\overset{N/4-1}{\mathrm{\Σ}}}{\left|r(n_{\mathrm{coarse}}+384+k)\right|}^2$ With

$P(n_{\mathrm{coarse}}+384)=\underset{k=0}{\overset{N/4-1}{\mathrm{\Σ}}}(n_{\mathrm{coarse}}+384-k)r(n_{\mathrm{coarse}}+384+k)$

Obtain: $M_{\mathrm{Pro}}(n_{\mathrm{coarse}}+384)=\frac{{\left|P(n_{\mathrm{coarse}}+384)\right|}^2}{{\left(R(n_{\mathrm{coarse}}+384)\right)}^2},$ Calculate then

R(n _coarse+385)＝R(n _coarse+384)-|r(n _coarse+384)| ²+|r(n _coarse+384+64)| ²

$P(n_{\mathrm{coarse}}+385)=\underset{k=0}{\overset{N/4-1}{\mathrm{\Σ}}}r(n_{\mathrm{coarse}}+385-k)r(n_{\mathrm{coarse}}+385+k)$

Obtain: $M_{\mathrm{Pro}}(n_{\mathrm{coarse}}+385)=\frac{{\left|P(n_{\mathrm{coarse}}+385)\right|}^2}{{\left(R(n_{\mathrm{coarse}}+385)\right)}^2}$

By that analogy, obtain one group of (such as 32) M _ProValue, the pairing sequence number n of maximum wherein _n Subtract 64 and equal n _Fine2The just smart sync bit n that will look for synchronously _Head2, the just position of first bit data behind the CP of frame head two, then the position n of first bit data behind frame head one CP _Head1Be n _Fine2-256-32=n _Fine2-288.

The another one benefit of the method is utilized exactly to grip altogether when symmetry is asked maximum correlation peak and not influenced by frequency deviation, and establishing and sending signal is x (n), neglects The noise, the baseband signal r of acceptance (n)=x (n) e ^{J2 π n ε/N}Can push away by (5) formula:

${\left|P\left(d\right)\right|}^2={\left|\underset{k=0}{\overset{N/4-1}{\mathrm{\&Sigma;}}}r(d-k)r(d+k)\right|}^2={\left|\underset{k=0}{\overset{N/4-1}{\mathrm{\&Sigma;}}}x(d-k){\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">e</figure-callout>}}^{\frac{j}{}}x(d+k){\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">e</figure-callout>}}^{\frac{j}{}}\right|}^2$

Following formula as seen, maximum correlation peak value function and frequency deviation are irrelevant, therefore smart timing can be placed on before the frequency offset estimating synchronously, not only solve decimal frequency bias and estimated the big problem of mean square error, and avoided regularly influencing each other with the frequency deviation mistake, make various synchronized algorithms more independent, greatly improve the robustness of system.

(3) decimal frequency bias is estimated

If the absolute frequency deviation of sending and receiving end is Δ f, the systematic sampling rate is f _s, N is a sub-carrier number, is defined as with respect to the frequency deviation coefficient of subcarrier spacing $\mathrm{\&epsiv;}=\frac{\mathrm{\&Delta;f}}{f_s/N},$ The length of each sequence in the repetitive sequence of D frame head, z is for postponing relevant slip function, the position n of first bit data behind frame head one CP _Head1, the position n of first bit data behind frame head two CP _Head2, the thick frequency deviation of decimal is ε _Coarse, the smart frequency deviation of decimal is ε _Fine

The decimal frequency bias estimator that the present invention adopts is

$\mathrm{\&epsiv;}=-\frac{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">N</figure-callout>}}{2\mathrm{\&pi;D}}\mathrm{angle}\left(z\right)---\left(8\right)$

If the transmission signal is x (n), neglect The noise, the baseband signal r of acceptance (n)=x (n) e ^{J2 π n ε/N}

$z=\underset{n=0}{\overset{D-1}{\mathrm{\&Sigma;}}}r\left(n\right)r^{*}(n+D)$

$=\underset{n=0}{\overset{D-1}{\mathrm{\&Sigma;}}}x\left(n\right)x^{*}(n+D){\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">e</figure-callout>}}^j{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;\&epsiv;}(n+D)/N}$

$=e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;\&epsiv;D}/N}\underset{n=0}{\overset{D-1}{\mathrm{\&Sigma;}}}{\left|x\left(n\right)\right|}^2---\left(9\right)$

Because angle (z) is defined on the interval [π π], therefore estimable frequency deviation region is

$\left|\mathrm{\&epsiv;}\right|\&le;\frac{N}{2D}---\left(10\right)$

For short training sequence, its D=64, therefore estimable maximum frequency deviation is 2; For long training sequence, its D=126, estimable maximum frequency deviation is 1.

Owing to calculated in synchronously thick $c\left(n\right)=\underset{k=0}{\overset{63}{\mathrm{\&Sigma;}}}r(n+k)r^{*}(n+k+64),$ N=1...1000 wherein, when then doing the decimal frequency bias rough estimate by frame head one, c (n) does not need to calculate again, directly quotes the thick value that has calculated in synchronously and estimates decimal frequency bias, and then the thick frequency deviation of decimal is:

${\mathrm{\&epsiv;}}_{\mathrm{coarse}}=-\frac{256}{2\mathrm{\&pi;}*64}\mathrm{angle}\left(c\left(n_{\mathrm{head}1}\right)\right)$

Proofread and correct all then and receive data.After the correction, just do the smart estimation of decimal frequency bias by frame head two, $z=\underset{k=0}{\overset{127}{\mathrm{\&Sigma;}}}r({\mathrm{<figure-callout\; id="2"\; label="n\; head"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{head}}+k)r^{*}({\mathrm{<figure-callout\; id="2"\; label="n\; head"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{head}}+k+128),$ Then the smart frequency deviation of decimal is:

${\mathrm{\&epsiv;}}_{\mathrm{fine}}=-\frac{256}{2\mathrm{\&pi;}*128}\mathrm{angle}\left(z\right)$

And then proofread and correct a secondary data.

(4) integer frequency bias is estimated

802.16a the carrier-frequency range of system is 2.4G ~ 11GHz, the maximum crystal oscillator error that standard allows is 20ppm, if the clock of transmitter and receiver all is at the worst error state, and positive and negative opposite words, then the error between the transmitter and receiver is 40ppm, so possible maximum frequency deviation scope is:

ΔF＝40*10 ^-6*(2.4~11)*10 ⁹＝96~440KHz

And subcarrier spacing is:

Δf＝(1.75*m*10 ⁶)/256＝6.8359*m?KHz

So the maximum frequency deviation coefficient can reach about 65, only be that decimal frequency bias is corrected inadequately, therefore also to carry out integer frequency bias and correct.

The present invention adopts frame head two and frame head one frequency-domain combined estimation integer frequency bias.Because ICI has been avoided in the correction of decimal frequency bias, thereby frequency domain symbol is only because integer frequency bias produces displacement, and frame head one is consistent with the displacement of frame head two, and mobile figure place is 4 multiple; This method utilizes the ratio on frame head two and the frame head one frequency domain 4 gall nut carrier waves to form new known PN sequence v _k, with the corresponding PN sequence that receives do relevant, 4 sampled points search of each displacement maximums in possible integer frequency bias scope, the carry digit when determining maximum is the integer frequency bias coefficient.If a PN preface X={-W ,-W+4 ... ,-4,0,4 ..., W-4, W}, W are designated as 4 number of multiples, Y under this PN sequence satisfies _{1, n}N data of the first frequency domain of representative frame, Y _{2, n}Represent n data of frame head two frequency domains, v _kFor utilizing the ratio on frame head two and the frame head one frequency domain 4 gall nut carrier waves to form new known PN sequence.The integer frequency bias estimation function is B (g):

$B\left(g\right)=\frac{{\left|{\mathrm{\&Sigma;}}_{k\&Element;\mathrm{<figure-callout\; id="1"\; label="X\; Y"\; filenames="A20041004147800151.png,A20041004147800171.png"\; state="\{\{state\}\}">X</figure-callout>}}{Y}_{}^{}{}_{1,k+4g}^{*}{v}_k^{*}{\mathrm{<figure-callout\; id="2"\; label="Y"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">Y</figure-callout>}}_{2,k+4g}\right|}^2}{2{\left({\mathrm{\&Sigma;}}_{k\&Element;X}{\left|{\mathrm{<figure-callout\; id="2"\; label="Y"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">Y</figure-callout>}}_{2,k}\right|}^2\right)}^2}---\left(11\right)$

Concrete steps are as follows: according to calculating the sync bit that obtains synchronously, frame head is to received signal done the fft conversion, obtains the frequency-region signal Y of frame head earlier _{1, n}And Y _{2, n}, calculate then $v_k=\frac{{\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="A20041004147800151.png,A20041004147800171.png"\; state="\{\{state\}\}">Y</figure-callout>}}_{1,\mathrm{<figure-callout\; id="2"\; label="k\; Y"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">k</figure-callout>}}}{Y_{}}(k\&Element;X).$ Then in suitable scope,, calculate such as-10 ~ 10 when frequency deviation region (just search be-40 ~ 40) $B\left(g\right)=\frac{{\left|{\mathrm{\&Sigma;}}_{k\&Element;\mathrm{<figure-callout\; id="1"\; label="X\; Y"\; filenames="A20041004147800151.png,A20041004147800171.png"\; state="\{\{state\}\}">X</figure-callout>}}{Y}_{}^{}{}_{1,k+4g}^{*}{v}_k^{*}{\mathrm{<figure-callout\; id="2"\; label="Y"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">Y</figure-callout>}}_{2,k+4g}\right|}^2}{2{\left({\mathrm{\&Sigma;}}_{k\&Element;X}{\left|{\mathrm{<figure-callout\; id="2"\; label="Y"\; filenames="A20041004147800171.png"\; state="\{\{state\}\}">Y</figure-callout>}}_{2,k}\right|}^2\right)}^2},$ Get the wherein pairing sequence number g of B (g) the maximum, multiply by 4 again, be integer frequency bias.

(5) follow the tracks of synchronously

The OFDM symbol has been introduced Cyclic Prefix as protection at interval in order to resist multipath effect.Sample value in the Cyclic Prefix is identical with one section isometric sample value of symbol end, this ofdm signal intrinsic information redundancy be characterized as be realize symbol regularly and Frequency Synchronization provide possible.The present invention adopts maximum likelihood estimator module to carry out uniting of symbol timing and frequency deviation estimated to follow the tracks of in time domain, and the maximum likelihood estimator module structure as shown in Figure 7.

If N is the number of the data in the OFDM symbol, γ (θ) is that received signal N position postpones correlation function, and ζ (θ) is the received signal energy function, L _CpBe the length of cyclic prefix CP, SNR is a signal to noise ratio, $\mathrm{\&rho;}=\frac{\mathrm{SNR}}{\mathrm{SNR}+1},$ The position n of first bit data behind frame head one CP _Head1, the position n of first bit data behind frame head two CP _Head2, in addition for narrate make things convenient for defined variable α (θ)=| γ (θ) |-ρ ζ (θ).The symbol of tracking phase is θ regularly _MLWith frequency offset estimating ε _MLFor:

${\mathrm{\&theta;}}_{\mathrm{ML}}=\arg \underset{\mathrm{\&theta;}}{\max }\left(\right|\mathrm{\&gamma;}\left(\mathrm{\&theta;}\right)|-\mathrm{\&rho;\&zeta;}\left(\mathrm{\&theta;}\right))---\left(12\right)$

${\mathrm{\&epsiv;}}_{\mathrm{ML}}=-\frac{1}{2\mathrm{\&pi;}}\&angle;\mathrm{\&gamma;}\left({\mathrm{\&theta;}}_{\mathrm{ML}}\right)---\left(13\right)$

Wherein, $\mathrm{\&gamma;}\left(\mathrm{\&theta;}\right)=\underset{k=\mathrm{\&theta;}}{\overset{\mathrm{\&theta;}+L_{\mathrm{cp}}-1}{\mathrm{\&Sigma;}}}r\left(k\right)r^{*}(k+N)---\left(14\right)$

$\mathrm{\&zeta;}\left(\mathrm{\&theta;}\right)=\frac{1}{2}\underset{k=\mathrm{\&theta;}}{\overset{\mathrm{\&theta;}+L_{\mathrm{cp}}-1}{\mathrm{\&Sigma;}}}({\left|r\left(k\right)\right|}^2+{\left|r(k+N)\right|}^2)---\left(15\right)$

Regularly synchronized algorithm is not influenced by frequency deviation, and need not to transmit special synchronizing symbol, and efficiency of transmission is than higher; The amount of calculation of algorithm requires less, is easy to realize.

Concrete steps are as shown in Figure 8: according to the sync bit that obtains synchronously, find the sequence number n of first bit data behind the frame head _Data=n _Head1+ 288*2 is because the sequence number of first bit data may compare n behind the actual frame head _DataGreatly, also may be littler than the sequence number of first bit data behind the frame head, so at n _DataAbout seek sync bit more accurately, such as at n _DataIn ± 32 the scope, so calculate earlier

$\mathrm{\&gamma;}(-32)=\underset{k=-32}{\overset{-32+L_{\mathrm{cp}}-1}{\mathrm{\&Sigma;}}}r(n_{\mathrm{data}}+k)r^{*}(n_{\mathrm{data}}+k+256)$

$\mathrm{\&zeta;}(-32)=\frac{1}{2}\underset{k=-32}{\overset{-32+L_{\mathrm{cp}}-1}{\mathrm{\&Sigma;}}}({\left|r(n_{\mathrm{data}}+k)\right|}^2+{\left|r(n_{\mathrm{data}}+k+256)\right|}^2)$

α(-32)＝|γ(-32)|-ρζ(-32)

Calculate according to the thick iterative formula of mentioning in synchronously then

γ(-31)＝γ(-32)-r(n _data-32)r ^*(n _data-32+256)+r(n _data)r ^*(n _data+256)

$\mathrm{\&zeta;}(-31)=\mathrm{\&zeta;}(-32)-\frac{1}{2}({\left|r(n_{\mathrm{data}}-32)\right|}^2+{\left|r(n_{\mathrm{data}}-32+256)\right|}^2)$

$+\frac{1}{2}({\left|r\left(n_{\mathrm{data}}\right)\right|}^2+{\left|r(n_{\mathrm{data}}+256)\right|}^2)$

α(-31)＝|γ(-31)|-σζ(-31)

By that analogy, obtain the value of one group of α (θ), maximum corresponding sequence number θ wherein is the position θ of a data sign-on _MLThen ${\mathrm{\&epsiv;}}_{\mathrm{ML}}=-\frac{1}{2\mathrm{\&pi;}}\&angle;\mathrm{\&gamma;}\left({\mathrm{\&theta;}}_{\mathrm{ML}}\right),$ Then all bit data in this symbol are carried out compensate of frequency deviation.In like manner the 3rd symbol of second symbol behind frame head or the like done same synchronous tracking.

Claims (7)

1, the method for synchronous in a kind of OFDM BWA is characterized in that synchronous step is:

A, utilize frame head one to do to postpone relevant judgement, realize synchronously thick;

B, utilize frame head two grip symmetry altogether, realize regularly smart;

C, carry out the decimal frequency bias rough estimate, carry out with frame head two that decimal frequency bias is smart to be estimated with frame head one;

D, usefulness frame head two are estimated integer frequency bias with the frequency domain sequence of frame head one;

E, last tracking synchronously;

Wherein: frame head has comprised long training sequence and short training sequence, frame head one comprises the cycle of 4 repetitions for short training sequence, and frame head two comprises 2 repetition periods for long training sequence, this repetition period property generates by inserted 0 before the IFFT conversion, frame head two not only has cycle repeatability, and has and grip symmetry altogether.

2, the method for synchronous in the OFDM BWA according to claim 1, it is characterized in that thick synchronous method is: the periodicity of utilizing frame head one, adopt and postpone correlation technique, two sliding windows are arranged: postpone relevant sliding window C and signal energy computation sliding window P, one is the correlation function that calculates received signal and its inhibit signal, and retardation coefficient equals the Cycle Length 64 of Short Training symbol; The signal energy that another calculates in associated window is to be used for judgment variables is carried out normalized, and judgment variables just is no longer dependent on the abswolute level of energy like this.

3, the method for synchronous in the OFDM BWA according to claim 1, it is characterized in that smart method regularly is: that utilizes frame head two grips symmetry altogether, that asks the first half frame head and half frame head of back grips correlation function altogether as judgment variables, search for the maximum of this judgment variables, this maximum place is smart place, the just accurate starting point of a frame synchronously.

4, the method for synchronous in the OFDM BWA according to claim 1 is characterized in that the decimal frequency bias estimation approach is: carry out the decimal frequency bias rough estimate with frame head one, carry out the smart estimation of decimal frequency bias with frame head two.For short training sequence, estimable maximum frequency deviation is 2; For long training sequence, its time delay D=128, estimable maximum frequency deviation is 1.

5, the method for synchronous in the OFDM BWA according to claim 1, it is characterized in that the integer frequency bias estimation approach is: adopt frame head two and frame head one frequency-domain combined estimation integer frequency bias, this method utilizes the ratio on frame head two and the frame head one frequency domain 4 gall nut carrier waves to form new known pseudo-random noise sequence PN sequence v _k, with the corresponding PN sequence that receives do relevant, 4 sampled points search of each displacement maximums in possible integer frequency bias scope, the carry digit when determining maximum is the integer frequency bias coefficient.

6, the method for synchronous in the OFDM BWA according to claim 1, it is characterized in that the method for following the tracks of synchronously by: utilize the feature of information redundancy in the Cyclic Prefix intrinsic in the ofdm signal, adopt maximum likelihood estimator module time domain carry out symbol regularly and frequency deviation unite estimation.

7, the method for synchronous in the OFDM BWA according to claim 1, it is as follows to it is characterized in that postponing relevant decision method: ask a signal and its to pass through the normalization correlation function of the later signal of time delay as judgment variables, two sliding windows are arranged, one is to postpone relevant sliding window C, calculate the correlation function of received signal and its inhibit signal, retardation coefficient equals the Cycle Length 64 of Short Training symbol; Another is signal calculated energy sliding window P, calculates the signal energy in the associated window, is to be used for judgment variables is carried out normalized, and judgment variables just is no longer dependent on the abswolute level of energy like this; When judgment variables during, this place can be defined as just place synchronously greater than threshold value.

Images