CN1917491A - OFDM synchronization method of using training circulated prefix - Google Patents

Abstract

The method comprises: inserting the training sequence after the OFDM symbol to generate the cyclic prefix; fixing the training sequence segment in order to search a fitted timing offset; according to the timing offset, calculating the fractional frequency offset to get the integer frequency offset. The integer frequency offset is calculated using the training sequence section; by fixing the training sequence part in OFDM symbol, the value of timing bias is constantly changed to get a section of possible cyclic prefix of OFDM symbol, by which the integer frequency offset estimation, can be made so as to improve the estimation accuracy.

Description

Utilize the OFDM method for synchronous of training circulated prefix

Technical field

The present invention relates to wireless communication technology, relate in particular to the ofdm system in the radio communication.

Technical background

Orthogonal frequency division multiplex OFDM is the high speed transmission technology under a kind of wireless environment, and its feature is that each subcarrier is mutually orthogonal, makes the frequency spectrum after the band spectrum modulation can be overlapped, thereby has reduced the phase mutual interference between subcarrier.Yet with respect to single-carrier system, ofdm system is very responsive to timing and frequency shift (FS).In order to eliminate intersymbol interference (ISI) and inter-carrier interference (ICI), estimation synchronously just seems extremely important.

What prior art all adopted the synchronous estimation of ofdm system is about auxiliary (data-aided) algorithm for estimating of data and the auxiliary algorithm for estimating of non-data.Data are auxiliary to be estimated to have brought the loss of information rate owing to will cause the wasting of resources with a plurality of OFDM piece conduction frequency symbols or training sequence; Estimate there is not the wasting of resources though non-data are auxiliary, estimation performance is not good when multidiameter fading channel, and precision is not high.

The classical ML that utilizes data dependence character (maximum likelihood) algorithm is based on and puts forward under the Gaussian channel, but under multidiameter fading channel because the influence of channel, the repetitive nature variation of Cyclic Prefix makes that the ML functional value of regularly putting d and its both sides sample point is very approaching, causes the timing estimation performance seriously to descend.Because the performance of timing estimation is not good, make the performance of frequency offset estimating also descend thereupon.As [Van de Beek J J such as Van de BeekJ J, Sandell M, Boriesson P O.ML estimation of timeand frequency offset in OFDM systems.IEEE Transaction on SignalProcessing, 1997,45 (7): 1800-1805] principle relevant with the Cyclic Prefix data according to the data of OFDM symbol rear end, adopt the ML algorithm to carry out timing and frequency offset estimating simultaneously, but in fading channel poor effect of following time, and can not carry out integer frequency offset and estimate, so that synchronous effect is not good.

Summary of the invention

The present invention proposes a kind of OFDM method for synchronous that utilizes training circulated prefix, carry out the above-mentioned defective of frequency offset estimating and proposed a kind of new frequency deviation estimating method, remedied the above-mentioned deficiency of prior art at available technology adopting ML algorithm.The present invention solves the problems of the technologies described above the technical scheme that is adopted: propose a kind of orthogonal frequency division multiplex OFDM method for synchronous that utilizes training circulated prefix, the method comprising the steps of: insert training sequence in OFDM symbol rear end and produce Cyclic Prefix thus, training sequence section in the fixed OFDM symbol is also sought the timing offset value that is complementary with it, calculate fractional part of frequency offset according to the timing slip value, obtain the integer multiple frequency deviation then, realize that OFDM is synchronous.And by the training sequence part in the fixed OFDM symbol, constantly change the timing offset value and obtain one section Cyclic Prefix that the OFDM symbol is possible, improved symbol precision regularly, and can carry out integer frequency offset and estimate.Have only a kind of relation of cyclic shift between the two according to the training sequence behind training sequence before the inverse Fourier transform IFFT and the Fourier transform FFT, obtain the integer multiple frequency deviation according to correlativity between them matter.

Beneficial effect of the present invention, the OFDM synchronization method that utilizes training circulated prefix that adopts the present invention to propose, the timing estimation precision can significantly be promoted, and then improve the fractional part of frequency offset estimation performance, and can not carry out the defective that integer frequency offset is estimated with respect to the ML algorithm, the present invention has also well solved this problem.

Description of drawings

Fig. 1 utilizes the synchronous system model figure of the OFDM of training circulated prefix

Fig. 2 inserts the OFDM symbolic construction of training sequence

Fig. 3 adopts the timing estimation impact of performance comparison diagram of ML algorithm and the method for the invention

Embodiment

The present invention proposes simply a kind of and utilize the OFDM synchronized algorithm of training circulated prefix efficiently.Insert training sequence in OFDM symbol rear end, training sequence produces Cyclic Prefix thus, training sequence section in the fixed OFDM symbol is also sought the timing offset value that is complementary with it, do timing estimation at receiving terminal according to both relativity of time domain matter, and then carry out fractional part of frequency offset and estimate, after timing and fractional part of frequency offset correction, remove Cyclic Prefix and Fourier transform FFT, at this moment the frequency domain correlation matter according to original training sequence and received training sequence can obtain the integer frequency offset estimation, thereby realizes that OFDM is synchronous.

(1) sets up the OFDM synchronized algorithm system model that utilizes training circulated prefix

Be illustrated in figure 1 as the synchro system of the OFDM foundation that utilizes training circulated prefix, in ofdm system, the N of transmission ends complex symbol (d ₀D _kD _K-1) be to be modulated onto on N the subcarrier behind the OFDM symbol process inverse Fourier transform IFFT, back L (Tx) getting N complex symbol is as training sequence.This last L sample value (being training sequence) wherein is inserted in the beginning of each OFDM symbol to form protection at interval.After through parallel serial conversion P/S and inverse Fourier transform IFFT, baseband modulation signal s (n) is represented as:

$s\left(n\right)=\frac{1}{N}\underset{k=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{d}_k\exp \left(\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20061005447900101.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;k}(n-L)}{N}\right)$

The d here _kBe through the data symbol after the constellation point mapping.Each OFDM cycle is M=N+L.After s (n) was the multipath channel of h (n, 1) by impulse response, output signal x (n) was represented as:

$x\left(n\right)=\underset{l}{\mathrm{\&Sigma;}}h(n,l)s(n-l)$

At receiving terminal, because the influence of symbol timing offset and carrier frequency offset, actual reception signal r (n) is represented as:

r(n)＝x(n-d)e ^j2πΔfn/N+w(n)

Wherein d represents the symbol timing offset, and Δ f represents by the different of frequency of oscillation between receiving terminal and the transmitting terminal and carrier frequency offset that Doppler frequency shift causes.W (n) is the white Gaussian noise of zero-mean, and it and signal are separate.

At the time domain receiving terminal, taking out a segment length from received signal r (n) is the data segment of L

R(i)＝[r(i)，r+(i+1)，…，r(i+L-1)] i＝0，1，2，…

Wherein i is a timing offset, and when value i changed, the data segment R (i) that obtains was also different.

If the training sequence part on time domain is train (n), 0≤n≤L-1 can obtain the timing offset Δ d relevant with the fixed training sequence section by data dependence matter:

$\mathrm{\&Delta;d}=\arg m\underset{i}{a}x\left\{\mathrm{sum}\left(\right|R\left(i\right)*\mathrm{conj}\left(\mathrm{train}\right)\left|\right)\right\}$

After estimating the timing offset Δ d relevant, can be in the hope of the fractional part of frequency offset estimated value with the fixed training sequence section:

$\mathrm{\&Delta;}{f}_F=-\frac{1}{2\mathrm{\&pi;}}\&angle;\underset{k=\mathrm{\&Delta;d}}{\overset{\mathrm{\&Delta;d}+L-1}{\mathrm{\&Sigma;}}}r\left(k\right)r*(k+N)$

Through having determined the timing offset and the fractional part of frequency offset of system after timing and the fractional part of frequency offset correction, after removing Cyclic Prefix S/P and Fourier transform FFT, also remaining in the system has integer frequency offset.Training sequence in the OFDM symbol of this time-frequency domain is expressed as Z (k), k=0, and 1 ... L '-1 (wherein L ' is a training sequence length).According to the character of FFT computing, be equivalent to a displacement on frequency domain in the influence of frequency deviation on the time domain, we can obtain integer frequency offset and are estimated as thus:

$\mathrm{\&Delta;}{f}_I=\arg m\underset{j}{a}x\left\{\right|\underset{k=0}{\overset{L-1}{\mathrm{\&Sigma;}}}Z\left(k\right)d*{(k+j)}_{L^{\&prime;}}\left|\right\}-\frac{L^{\&prime;}}{2}\&le;\mathrm{\&Delta;}{f}_I\&le;\frac{L^{\&prime;}}{2}$

The d here (k+j) _{L '}Expression d (k) length is the cyclic shift of L '.

By following formula as can be known, the estimated value of integer multiple frequency deviation depends on the complex symbol d (k) of the preceding transmission ends of inverse Fourier transform IFFT and the training sequence Z (k) behind the Fourier transform FFT.

As shown in Figure 2, we insert training sequence Tx in the rear end of each OFDM symbol, and the front end that this training sequence is copied to OFDM symbol Tifft forms Cyclic Prefix Tg.At the time domain receiving terminal, received signal r (n) is through behind the serial/parallel conversion S/P, from parallel signal r ₀R _NR _N+L-1In take out a segment length and make related operation with training sequence part [train (n)] fixing on time domain for the data segment R (i) of L, by correlativity between them matter, timing slip Δ d when can be in the hope of correlation maximum.When value i changes, data segment R (i) is also different, constantly change timing offset value i and obtain one section Cyclic Prefix that the OFDM symbol is possible, if training sequence part on time domain is train (n), by correlativity between them matter, the size of timing slip i when can be in the hope of correlation maximum, obtain best Cyclic Prefix Tg thus, according to the timing offset value of trying to achieve,, try to achieve fractional part of frequency offset again according to ML algorithm or other algorithms.And then behind past Cyclic Prefix S/P and Fourier transform FFT, obtain integer frequency offset.

Compare with the ML algorithm, this method has been fixed the data segment (known training sequence) of OFDM symbol rear end, by changing the size of possible timing offset value i, find out one piece of data R (i) with fixed data section correlation maximum, trying to achieve timing offset Δ d compares with the ML algorithm, this method has been fixed the data (known training sequence) of OFDM symbol rear end, by changing the size of possible timing offset value i, find out the one piece of data R (i) with fixed data section correlation maximum, try to achieve timing offset Δ d.Therefore, the timing estimation precision of this algorithm can significantly be promoted, and then improves the fractional part of frequency offset estimation performance, and can not carry out the defective that integer frequency offset is estimated with respect to the ML algorithm, and this algorithm has also well solved this problem.Thereby better realize OFDM synchronously.

(2) examples of properties

We are under the TD-SCDMA system environments, and the OFDM major parameter of choosing is: sampling frequency F=1.6MHz; Total sub-carrier number N=256; Average mark is given 16 users, 16 subcarriers of each CU; 16 subcarriers of last 1 user are put the training sequence midamble sign indicating number in the TD-SCDMA system; 4 OFDM symbols of the TD-SCDMA Time Slot Occupancy of 1 0.675ms; Cyclic Prefix takies 14 sample points.Concrete OFDM parameter setting sees Table 1.Data symbol adopts the QPSK modulation, training sequence adopts the basic midamble sign indicating number in the TD-SCDMA system, 128 original midamble sign indicating numbers are handled the back through scrambling use preceding 64, average 4 the OFDM symbols interior (1 TD-SCDMA time slot uses 4 OFDM symbols) that insert, each OFDM symbol uses 16, forms Cyclic Prefix at back 14 that get it.

The major parameter of table 1 ofdm system

Parameter	Parameter value
		Sampling frequency	16MHz
The OFDM symbolic number of a Time Slot Occupancy	4
		Total sub-carrier number	256
Midamble numeral carrier number	16
		The OFDM mark space	168.75μs(270chip)
FFT period T fft	160μs(256chip)
		Circulating prefix-length (protection at interval)	8.75μs(14chip)
Sub-carrier frequencies interval delta f	6.25kHz(1.6MHz/256)
		Number of users	16
The sub-carrier number that the user uses	16

Under the Case3 channel circumstance, the method that adopts ML algorithm and the present invention to propose is compared, from accompanying drawing 3 as can be seen, the possibility that adopts method of the present invention correctly to estimate the timing slip position exceeds nearly 20% than ML algorithm, the timing estimation performance increases substantially, and owing to increasing substantially of timing estimation performance, and then makes that the fractional part of frequency offset estimated performance is also significantly improved, and can carry out integer frequency offset and estimate, remedy and adopted the ML algorithm to carry out the deficiency of frequency offset estimating.Can also satisfy the requirement of TD-SCDMA system frame structure, when system is in slow fading channel, can utilize in the time slot a plurality of OFDM symbols to estimate simultaneously.Therefore, this algorithm can solve the OFDM stationary problem in the reality fully.

Claims (4)

1. orthogonal frequency division multiplex OFDM method for synchronous that utilizes training circulated prefix, insert training sequence in OFDM symbol rear end and produce Cyclic Prefix thus, it is characterized in that, training sequence section in the fixed OFDM symbol is also sought the timing offset value that is complementary with it, change the timing offset value and obtain possible Cyclic Prefix, calculate decimal overtones band deviation according to the timing slip value, obtain the integer multiple frequency deviation then, realize that OFDM is synchronous.

2. orthogonal frequency division multiplex OFDM method for synchronous according to claim 1 is characterized in that, adopts maximum likelihood algorithm ML to calculate fractional part of frequency offset, goes to obtain integer frequency offset behind Cyclic Prefix and the Fourier transform FFT.

3. orthogonal frequency division multiplex OFDM method for synchronous according to claim 1 is characterized in that, constantly changes the Cyclic Prefix that the timing offset value obtains one section OFDM symbol the best.

4. orthogonal frequency division multiplex OFDM method for synchronous according to claim 1 is characterized in that, according to the training sequence computes integer overtones band deviation behind complex symbol before the inverse Fourier transform IFFT and the Fourier transform FFT.

Images