CN105024966A - Symbol timing synchronization method of OFDM system - Google Patents

Abstract

The invention discloses a symbol timing synchronization method of an OFDM system. With the method, problems that no single sharp timing offset estimation function peak value is provided in the prior art and too much time is spent in determining an ideal timing position can be solved. The method comprises the following steps: 1, generating a training sequence and inserting the training sequence in front of a load data packet; 2, calculating a timing offset estimation function value at an initial position time of a sliding window at a receiving terminal; 3, taking an average of timing offset estimation functions of M times before the time at any time; 4, determining a weighting coefficient of the timing offset estimation function average value; 5, calculating a dynamic threshold according to the obtained timing offset estimation function average value and weighting coefficient; and 6, comparing the timing offset estimation functions and the dynamic threshold until detecting an ideal timing position of the current load data packet; and carrying out the operation continuously according to the mechanism to obtain an ideal timing position of each subsequent data packet. Therefore, timing can be realized accurately in a low-signal-to-noise-ratio environment and the real-time performance is good.

Description

The symbol timing synchronization method of ofdm system

Technical field

The invention belongs to communication technical field, particularly relate to a kind of symbol timing synchronization method, can be used for carrying out real-time timing synchronization in ofdm system

Background technology

Orthogonal frequency division multiplex OFDM technology is a kind of Multicarrier Transmission Technology, it by serial to parallel conversion by data flow lower for data flow decomposition rate at a high speed, then mutually orthogonal subcarrier is modulated into, overlap each other between each subcarrier spectrum, utilize these advantages effectively to improve the availability of frequency spectrum and to resist channel disturbance.

Although OFDM technology has a lot of advantage but compared with single carrier, the application of multi-transceiver technology makes it very harsh to the requirement of Domain Synchronous and frequency domain synchronization, if it is synchronous to set up good time-frequency domain, system will be subject to serious interference, cannot obtain correct demodulation result.Timing Synchronization, for determining the original position of FFT demodulation window, therefore must obtain good timing synchronization performance to obtain good time-frequency domain net synchronization capability.

The algorithm carrying out OFDM symbol timing at present mainly contains two classes: based on the signal-timing method of training sequence additional type and the signal-timing method of non-training sequence additional type.Signal-timing method based on training sequence utilizes the distinctive structure of sequence and correlation to obtain timing slip estimation function, although this method is lower than non-training sequence auxiliary type method in the availability of frequency spectrum, but more accurate Symbol Timing result can be obtained, and still can obtain good performance under severe channel circumstance.The classical signal-timing method based on training sequence has Schmidl method, Minn method and Park method, these three kinds of algorithms utilize the design feature of training sequence to adopt corresponding related operation mode to obtain timing slip estimation function, obtain timing slip estimation function peak value by getting maximum, and then determine the original position of training sequence and desirable Symbol Timing position.

But these algorithms but exist obvious problem when practice, mainly contain following two aspects:

One is spike problem: Schmidl method has flat peak phenomenon when there being Cyclic Prefix, decline near peak value when there is no Cyclic Prefix mild, also there is peak value not acute problem in same Minn method, and timing slip estimation function also there will be more secondary lobe, although Park method there will be sharp-pointed main peak also can with two sharp-pointed submaximums.

Two is real time problems: Schmidl method, Minn method and Park method are all that the maximum by obtaining timing slip estimation function locks timing position, obtain maximum just to need to compare continuously in a symbol, cannot judge when namely obtaining a timing slip estimation function that it is peak value actually immediately, judge that a timing slip estimation function maximum at least needs an OFDM symbol cycle of wait to determine, this adds increased Symbol Timing process processing delay, cannot be real-time complete timing synchronization.

Summary of the invention

The object of the invention is to the symbol timing synchronization method proposing a kind of ofdm system, to solve above-mentioned prior art without single sharp-pointed timing slip estimation function peak value, and obtain peak value determination timing position long problem consuming time.

Technical scheme of the present invention is achieved in that

(1) before the load data bag of ofdm system, inserting a length is the training sequence of N, and N is the subcarrier number of ofdm system;

(2) at the receiving terminal of ofdm system, the sliding window utilizing length to be N is sampled to reception data, calculates timing slip estimation function value F (m) in sliding window original position m moment;

(3) dynamic threshold in m moment is calculated: $G\left(m\right)=a\·\frac{1}{M}\underset{k=-M}{\overset{-1}{\Σ}}F(m+k),$

Wherein, m > M, a are constants, and its size is obtained by emulation testing; M be greater than 100 a constant, F (m+k) represents the timing slip estimation function value in m+k moment ,-M≤k≤-1;

(4) timing slip estimation function value F (m) in m moment and dynamic threshold G (m) are compared, determine current load data bag D _iideal symbol timing position if F (m)≤G (m), then can't detect current load data bag D _iideal timing position, perform step (5); If F (m) > G (m), then perform step (6);

(5) make m=m+1, upgrade F (m) and G (m), repeat step (4) until current load data bag D detected _iideal timing position complete current load data bag D _itiming synchronization;

(6) make m=m+1, upgrade F (m) and G (m), repeat step (4), to next load data bag D _i+1ideal timing position detect, complete next load data bag D _i+1timing synchronization.

The symbol timing synchronization method of ofdm system provided by the invention, has the following advantages:

1, under low signal-to-noise ratio environment, timing is accurately stable, function admirable:

At transmitting terminal, the training sequence inserted before load data bag is generated by CAZAC sequence, because CAZAC sequence all has good autocorrelation at time-frequency domain, therefore training sequence has good autocorrelation, receiving terminal can be made to calculate the timing slip estimation function peak value obtained more sharp-pointed;

At receiving terminal, related operation is carried out to the sampled data entering sliding window, because the related operation mode in sliding window designs according to the design feature of training sequence, can make to obtain timing slip estimation function peak value and only occur one in a load data bag, make to ensure that under low signal-to-noise ratio environment timing is accurately stable and performance is good.

2, real-time

The present invention to be compared with dynamic threshold by timing slip estimation function value and determines ideal timing position, because the dynamic threshold of any time is only relevant with the timing slip estimation function value produced before this moment, the dynamic threshold in this moment therefore can be tried to achieve in real time;

Judge that bias estimation peak of function determination ideal timing position only needs timing slip estimation function value and dynamic threshold once to compare, three kinds of classic algorithm Schmidl, Minn, Park methods then need to compare continuously, need to wait for that long time could judge bias estimation peak of function determination ideal timing position, the stand-by period of testing process can be greatly reduced relative to three kinds of classic algorithm the present invention, there is stronger real-time.

Accompanying drawing explanation

Fig. 1 of the present inventionly realizes general flow chart;

Fig. 2 is that the present invention enters the schematic diagram of sliding window three kinds of situations at receiving terminal training sequence;

Fig. 3 be the present invention obtain timing slip estimation function average and dynamic threshold realize block diagram;

Fig. 4 is the relativeness schematic diagram of the present invention's timing slip function and dynamic threshold when choosing different weights constant coefficient a in SNR=-4dB situation;

Fig. 5 is that the present invention detects the sub-process figure of ideal timing position according to timing slip estimation function and dynamic threshold;

Fig. 6 is the curve chart that the metrical error mean-square value of existing Schmidl, Minn, Park tri-kinds of classical signal-timing methods changes with signal to noise ratio snr.

Embodiment

Clearly understand to make application target of the present invention, Implementation Technology and advantage thereof, below in conjunction with drawings and Examples, content of the present invention is further described in detail, be to be understood that, specific embodiment described herein only in order to explain the present invention, is not limited to the present invention.

With reference to Fig. 1, specific implementation step of the present invention is as follows:

Step 1: generate training sequence, and before training sequence is inserted into load data bag.

Generating the existing method of training sequence has multiple, such as Schmidl method, Minn method, Park method, wherein:

Schmidl method, 0,1 sequence is adopted to modulate through BPSK/QPSK/QAM, generate sequence { D (0), D (1), D (2) that length is N/2, ... D (N/2-1) }, to sequence { D (0), D (1), D (2), ... D (N/2-1) } carry out N/2 point IFFT modulation, generate length be the time domain sequences of N/2 by time domain sequences carry out permutation and combination, obtain the training sequence of Schmidl method

In Minn method, 0,1 sequence is adopted to modulate through BPSK/QPSK/QAM, generate sequence { K (0), K (1), K (2) that length is N/4, ... K (N/4-1) }, to sequence { K (0), K (1), K (2), ... K (N/4-1) } carry out N/4 point IFFT modulation, generate length be the time domain sequences of N/4 by time domain sequences carry out permutation and combination, obtain the training sequence of Minn method

In Park method, 0,1 sequence is adopted to modulate through BPSK/QPSK/QAM, generate sequence { S (0), S (1), S (2) that length is N/4, ... S (N/4-1) }, to sequence { S (0), S (1), S (2), ... S (N/4-1) } carry out N/4 point IFFT modulation, generate length be the time domain sequences of N/4 conjugate symmetric sequence be $\stackrel{\→}{T}=\{s^{*}(N/4-1),s^{*}(N/4-2),s^{*}(N/4-3),\mathrm{...}{s}^{*}\left(0\right)\},$ By time domain sequences and sequence carry out permutation and combination, obtain the training sequence of Park method

This example adopts the structure of Park method to carry out the generation of training sequence, and its step is as follows:

(1a) computing formula of permanent envelope zero auto-correlation CAZAC sequence is utilized generate sequence that one group of length is N/4 C (0), C (1), C (2) ... C (N/4-1) }, wherein n=0,1 ..N/4-1;

Described CAZAC sequence is a kind of Zadoff-chu sequence, and computing formula is:

Wherein, K is the length of sequence, and W and K is relatively prime, and q is arbitrary integer.

The present invention's CAZAC sequence length used is K=N/4, N/4 is an even number, and get W=1, q=0, the computing formula that can obtain CAZAC sequence used in the present invention is thus n=0,1 ..N/4-1;

(1b) to sequence C (0), C (1), C (2) ... C (N/4-1) } carry out N/4 point IFFT modulation, generate length be the time domain sequences of N/4 conjugate symmetric sequence be $\stackrel{\→}{B}=\{c^{*}(N/4-1),c^{*}(N/4-2),c^{*}(N/4-3),\mathrm{...}{c}^{*}\left(0\right)\};$

(1c) by time domain sequences and sequence carry out permutation and combination, obtain training sequence

(1d), before training sequence being inserted into load data bag, namely before transmitting terminal sends a packet at every turn, first training sequence is sent.

Step 2: at receiving terminal, calculates timing slip estimation function value F (m) in sliding window original position m moment.

(2a) at receiving terminal, utilize sliding window to sample to received signal, and correlation p (m) that related operation obtains carried out to sampled data in sliding window:

$p\left(m\right)=2*\min \left(\right|\underset{k=0}{\overset{N/4-1}{\Σ}}r(m+k)\·r(m+N-k)|,|\underset{k=0}{\overset{N/4-1}{\Σ}}r(m+N/4+k)\·r(m+3N/4-k)\left|\right)$

Wherein min () gets minimum operation, and p (m) is namely correlation and correlation two times of middle minimum value, at 0≤k≤N/4-1, r (m+k), r (m+N-k), r (m+N/4+k), r (m+3N/4-k) all represent that in sliding window, sampling obtains four different pieces of informations;

In the Park method of classics, the computing formula of p (m) is:

$p\left(m\right)=|\underset{k=0}{\overset{N/2-1}{\Σ}}r(m+k)\·r(m+N-k)|$

At receiving terminal, if when the data of training sequence enter sliding window completely, as shown in Fig. 2 (a), p (m) value that p (m) value obtained by the inventive method and Park method obtain can be all a very large value, thus ensures that the timing slip function obtained has a sharp-pointed main peak; If the data of training sequence, when part enters sliding window, as shown in Fig. 2 (b) He Fig. 2 (c), the p (m) using Park method to calculate acquisition is also a larger value, but timing slip estimation function can be caused to occur a submaximum, correlation under both of these case with correlation all there is a value much larger than another value, and the present invention gets value less among the two to generate p (m), thus make p (m) become a less value, avoid the appearance of submaximum, make the timing slip estimation function peak value that obtains unique.

(2b) mean value of front 1/2 partial data gross power and rear 1/2 partial data gross power in sliding window is calculated:

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

Wherein, r (m+k), r (m+N-k) represent in sliding window two different pieces of informations obtained of sampling, 0≤k≤N/2-1;

(2c) timing slip estimation function value F (m) in m moment is calculated according to the result of (2a) and (2b):

$F\left(m\right)=\frac{|p\left(m\right){|}^2}{{\left(R\right(m\left)\right)}^2}.$

Step 3: in any m moment, is averaged the timing slip estimation function in m-M ~ m-1 moment, obtains average

With reference to Fig. 3, this step is by realizing by shift register and accumulator, to reduce calculating process.

(3a) the timing slip estimation function value newly produced in each moment enters shift register, and the data entering register input to accumulator simultaneously and carry out accumulation process;

(3b) M the data that former register stores are moved one by one afterwards, and last data that former register is stored shift out register, and the data shifting out register input to accumulator and carry out regressive process.

Because register initial storage data are 0 entirely, therefore starting the data that register shifts out is 0 entirely, only carry out accumulation process, when M+1 the data F (M+1) produced enters register, the value of accumulator be the timing slip estimation function value in 1 ~ M moment with therefore, when timing slip estimation function F (m) enters register, m > M, the result of accumulator is the summation of the timing slip estimation function in m-M ~ m-1 moment its value is multiplied by the average that coefficient 1/M can obtain the timing slip estimation function value in m-M ~ m-1 moment

Step 4: the weight coefficient a determining timing slip estimation function average.

Fig. 4 shows when signal to noise ratio snr=-4dB, the relativeness of timing slip function and dynamic threshold when choosing different weights coefficient a:

As weight coefficient a=5, dynamic threshold is smaller, not only be greater than dynamic threshold in ideal timing position timing slip estimation function value, there is the situation that timing slip estimation function is greater than dynamic threshold equally in other position, the timing position detected like this is wrong;

As weight coefficient a=15, only at ideal timing position, just have the situation that timing slip estimation function is greater than dynamic threshold, so just accurately can detect ideal timing position;

As weight coefficient a=30, all timing slip estimation function values are all less than dynamic threshold, cannot ideal timing position be detected like this.

It can thus be appreciated that a value is too small, System with Timing Errors position timing slip estimation function value is greater than dynamic threshold, occur that the position that namely false retrieval situation detects is not equal to ideal timing position, reduce the precision of detection, a value is excessive, be less than dynamic threshold in ideal timing position timing slip estimation function value, occur that namely undetected situation can't detect ideal timing position.Definition p _efor wrong probability of false detection, the timing position namely detected is not equal to the probability of ideal timing position; p _hfor false dismissal probability, be namely all less than the dynamic threshold corresponding with it at the timing slip estimation function of ideal timing position, can't detect the probability of ideal timing position; Probability p=1-p that ideal timing position correctly detects _e-p _h, most suitable a value can obtain according to following steps:

(4a) choose different b as alternative weight coefficient, perform step 5 and step 6, carry out ideal timing position detection, to simulate under different weights coefficient b p about the performance number of signal to noise ratio snr;

(4b) the minimum signal to noise ratio snr of p=100% is ensured under obtaining different b according to simulation performance _b;

(4c) minimum SNR is got _bcorresponding weight coefficient b as weight coefficient a, namely

Step 5: according to the timing slip estimation function average obtained dynamic threshold G (m) is calculated with weight coefficient a:

$G\left(m\right)=a\·\frac{1}{M}\underset{k=-M}{\overset{-1}{\Σ}}F(m+k).$

Step 6: the timing slip estimation function value and the dynamic threshold that compare acquisition, until detect the ideal timing position of current load data bag, by this machine-processed continuous firing with the ideal timing position of each packet after obtaining.

With reference to figure 5, being implemented as follows of this step:

(6a) timing slip estimation function value F (m) in m moment and dynamic threshold G (m) are compared, determine current load data bag D _iideal symbol timing position if F (m)≤G (m), then can't detect current load data bag D _iideal timing position, perform step (6b); If F (m) > G (m), then perform step (6c);

(6b) make m=m+1, upgrade F (m) and G (m), repeat step (6a) until current load data bag D detected _iideal timing position complete current load data bag D _itiming synchronization;

(6c) make m=m+1, upgrade F (m) and G (m), repeat step (6a), to next load data bag D _i+1ideal timing position detect, complete next load data bag D _i+1timing synchronization.

Effect of the present invention is further described by following emulation:

1, simulated conditions:

Choosing N is that 1024, M gets 200, and load data bag is an OFDM symbol, and length is L=1024, and simulated channel is additive Gaussian channel.

2, content and interpretation of result is emulated:

Emulation 1, when diplomatic copy invention weight coefficient a gets under different parameters, the situation that the correct verification and measurement ratio p of ideal timing position changes along with signal to noise ratio snr, simulation result is as shown in table 1:

Table 1

The minimum signal to noise ratio snr of ideal timing position correct verification and measurement ratio p=100% is ensured under can reading different coefficient a by table 1 _a, at all SNR _ain, SNR ₂₂=-3dB is minimum, and therefore weight coefficient a=22 is most suitable, and visible the present invention can obtain the correct verification and measurement ratio of ideal timing position of 100% under the signal to noise ratio condition of-3dB.

Emulation 2, the curve chart that the metrical error mean-square value emulating existing Schmidl, Minn, Park tri-kinds of classical signal-timing methods changes with signal to noise ratio snr, simulation result as shown in Figure 6:

As shown in Figure 6, it is 1dB ,-2dB ,-4dB that the signal to noise ratio lower limit that existing Schmidl, Minn, Park method detects under Gaussian channel divides.

From emulation 1 and emulation 2, the present invention, than classical Schmidl method and Minn method function admirable, only has the decay of 1dB compared with Park performance, therefore the present invention can under low signal-to-noise ratio environment accurate timing.

The inventive method determines that an ideal timing position only needs once to compare, need a sample period time, three kinds of classical algorithms need to compare continuously, need wait for that N+L=2048 sample period time could determine an ideal timing position, the inventive method determine that the time of an ideal timing position is only that classic algorithm determines an ideal timing position required time 0.049%, greatly reduce the stand-by period detected needed for ideal timing position, it can thus be appreciated that the present invention has very strong real-time.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (3)

1. The real-time symbol timing synchronization method of 1.OFDM system, is characterized in that:

   (1) before the load data bag of ofdm system, inserting a length is the training sequence of N, and N is the subcarrier number of ofdm system;

   (2) at the receiving terminal of ofdm system, the sliding window utilizing length to be N is sampled to reception data, calculates timing slip estimation function value F (m) in sliding window original position m moment;

   (3) dynamic threshold in m moment is calculated: wherein, m > M, a are constants, and its size is obtained by emulation testing; M be greater than 100 a constant, F (m+k) represents the timing slip estimation function value in m+k moment ,-M≤k≤-1;

   (4) timing slip estimation function value F (m) in m moment and dynamic threshold G (m) are compared, determine current load data bag D _iideal symbol timing position if F (m)≤G (m), then can't detect current load data bag D _iideal timing position, perform step (5); If F (m) > G (m), then perform step (6);

   (5) make m=m+1, upgrade F (m) and G (m), repeat step (4) until current load data bag D detected _iideal timing position complete current load data bag D _itiming synchronization;

   (6) make m=m+1, upgrade F (m) and G (m), repeat step (4), to next load data bag D _i+1ideal timing position detect, complete next load data bag D _i+1timing synchronization.
2. 2. method as claimed in claim 1, the training sequence in described step (1), obtains as follows:

   (1a) computing formula of permanent envelope zero auto-correlation CAZAC sequence is utilized generate sequence that one group of length is N/4 C (0), C (1), C (2) ... C (N/4-1) }, wherein n=0,1 ..N/4-1;

   (1b) to sequence C (0), C (1), C (2) ... C (N/4-1) } carry out N/4 point IFFT modulation, generate length be the time domain sequences of N/4 $\stackrel{\→}{A}=\{c\left(0\right),c\left(1\right),c\left(2\right),\mathrm{...}c(N/4-1)\},$ conjugate symmetric sequence be $\stackrel{\→}{B}=\{c^{*}(N/4-1),c^{*}(N/4-2),c^{*}(N/4-3),\mathrm{...}{c}^{*}\left(0\right)\};$

   (1c) by time domain sequences and sequence carry out permutation and combination, obtain training sequence
3. 3. method as claimed in claim 1, calculates timing slip estimation function value F (m) in sliding window original position m moment, carries out in accordance with the following steps in described step (2):

   (2a) related operation is carried out to data in sliding window, obtains correlation p (m):

   $p\left(m\right)=2*\min \left(\right|\underset{k=0}{\overset{N/4-1}{\Σ}}r(m+k)\·r(m+N-k)|,|\underset{k=0}{\overset{N/4-1}{\Σ}}r(m+N/4+k)\·r(m+3N/4-k)\left|\right)$

   Wherein min () gets minimum operation, and p (m) is namely correlation and correlation two times of middle minimum value, r (m+k), r (m+N-k), r (m+N/4+k), r (m+3N/4-k) represent that in sliding window, sampling obtains four different pieces of informations, 0≤k≤N/4-1;

   (2b) mean value of front 1/2 partial data gross power and rear 1/2 partial data gross power in sliding window is calculated:

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

   Wherein, r (m+k), r (m+N-k) represent in sliding window two different pieces of informations obtained of sampling, 0≤k≤N/2-1;

   (2c) timing slip estimation function value F (m) in m moment is calculated according to the result of (2a) and (2b):

   $F\left(m\right)=\frac{|p\left(m\right){|}^2}{{\left(R\right(m\left)\right)}^2}.$