CN101094517A

CN101094517A - Improved time synchronization method

Info

Publication number: CN101094517A
Application number: CNA2007100245140A
Authority: CN
Inventors: 徐大专; 王磊
Original assignee: Nanjing University of Aeronautics and Astronautics
Current assignee: Nanjing University of Aeronautics and Astronautics
Priority date: 2007-06-21
Filing date: 2007-06-21
Publication date: 2007-12-26
Anticipated expiration: 2027-06-21
Also published as: CN100574509C

Abstract

Regarding the influence of signal noise and signal offset to the sync performance, the invention provides a sync sequence design solution which divides the sync sequence into multi shorter-length segments; making correlated calculation for the received signals and each segment so as to generate multi correlation curves which are delayed and summarized to find the correlated peak and to complete the sync.

Description

A kind of improved method for synchronizing time

Technical field

The present invention relates to wireless communication field, particularly the time synchronization problem in wireless communication field.

Background technology

The time synchronized module is used for finding out the original position of useful signal in communication system, is the first step of system's reliability service.The channel quality of wireless channel is relatively poor relatively, not only exists serious multipath fading phenomenon and also has bigger noise jamming, and typical in the short wave communication channel, signal to noise ratio can be hanged down-3dB.In addition because the carrier frequency of transmitter and receiver and signals sampling rate can not be in full accord, cause also existing in the signal after the frequency conversion frequency shift (FS) to a certain degree.When utilizing local sequence synchronous, frequency shift (FS) is very big to the influence of net synchronization capability.

Method for synchronizing time commonly used mainly contains two classes:

(1) utilize the correlation of received signal self to carry out synchronously.Send the very strong sequence of two sections repetitions or correlation earlier as training sequence before sending signal, receiving terminal is relevant before and after input signal is done, and seeks relevant peaks.Such as ofdm system, can utilize the Cyclic Prefix of OFDM symbol and the repeatability of symbol itself to do synchronously.This homochronousness algorithm has been owing to utilized correlation before and after the signal self, offset fixedly frequency deviation to the influence of correlation.But all passed through channel owing to participate in two segment signals of related operation itself, a large amount of noise that superposeed is so under the relatively poor situation of channel quality, particularly in the remote-wireless channel, synchronization performance is poor.And because the existence of part correlation, the correlation curve that this method is tried to achieve is slowly to change, and gives to seek accurately relevant peaks and bring difficulty.

(2) utilize received signal and the correlation that is stored in local signal to carry out synchronously.Send one section specific training sequence earlier before sending signal, and sequence is stored in this locality, receiving terminal is made related operation with the training sequence of received signal and local storage, obtains the position of relevant peaks.This method is not because local sequence has superimposed noise, so even under the situation of low signal-to-noise ratio, can obtain bigger relevant peaks yet.But the method is to fixedly frequency deviation is very responsive.Frequency deviation can make the phase difference of received signal and local sequence along with the time becomes greatly gradually, thereby causes the correlation of synchronous head integral body sharply to descend, and when the excursion of phase difference reached 2 π, correlation disappeared substantially.

Summary of the invention

The present invention is directed to the pluses and minuses of the 2nd class methods, proposed a kind of improved time synchronized algorithm, and optimize at the operand of this algorithm.

Synchronous head partly is divided into original one section synchronizing sequence N isometric segment.It is thereby that received signal is done relevantly to find common relevant peaks to determine sync bit then with the short local sequence of N section respectively that receiving terminal is realized synchronous principle.Owing to be to make related operation, so still can overcome of the influence of bigger noise jamming to net synchronization capability with local sequence.Simultaneously every section local sequence length is all very short, so fixedly the excursion of the phase deviation that causes in every section of frequency deviation can be very not big, needs only excursion less than π, and every segment data and local sequence still can keep bigger correlation.

If but directly implement according to the method, received signal will be made the N correlations with local sequence, and operand can be quite big.Consider that doing relevant with the local sequence of N section all is same receiving sequence, so the employing shortcut calculation.Way is as follows:

At first store the frequency domain data of local sequence and the energy value of every segment data at receiving terminal.Be respectively that the local sequence of M is got conjugation and reversed with the N segment length, zero padding is done to preserve behind the FFT then.Wherein counter-rotating is for related operation equivalence is become convolution algorithm, and zero padding is to replace the requirement of linear convolution to sequence length in order to satisfy with circular convolution.Obtain the energy value of every section local sequence then, be stored in receiving terminal, be used for participating in the normalization computing.These work can be finished in advance, so do not need the consumption systems amount of calculation in when operation.

After receiving terminal receives signal, earlier signal is pressed the segmentation of L point.According to the thought of overlap-save method, with the L point data of each section and the last period the end the M point data be L+M point FFT (guarantee that L+M is 2 power, the 1st segment data is mended M zero in front) together.The result of FFT is multiplied each other with the N section frequency domain data of this locality respectively, at last multiplied result is remake IFFT, give up front M point, just obtained the correlation of N segment length for L.

Ask then input data each point square, obtain the corresponding energy value of importing data of each reference point with the method for recursion, will import the data energy and local sequence energy is asked geometric average, draw and be used for normalized energy value.Divided by energy, just obtain normalized correlation with the correlation of trying to achieve previously.

Ask the mean value of N section normalization correlation again, because N section synchronizing sequence sends successively, so peak value also occurs successively when relevant with the local sequence work of each section, so with different the counting of each section correlation time-delay, the 1st section correlation time-delay M* (N-1) point, the 2nd section time-delay M* (N-2) point, and the like, to ask average after the correlation addition after the time-delay, just can get correlation to the end.

Set thresholding at last, correlation is carried out peak value detect, find out the position of peak value, and then obtain the original position of synchronous head, the deadline is synchronous.

The invention has the advantages that: 1, this synchronized algorithm becomes the short segment of a lot of length with local sequences Design, both kept with local sequence do relevant can antimierophonic advantage, solved the sharply problem of decline of the whole correlation that draws because of frequency deviation again.2, for the related operation part of input signal and N segment data, the present invention is translated into 1 FFT computing and N IFFT computing, so just can utilize unified FFT module to improve the synchronous efficiency of system greatly.For the normalization computing of correlation, the present invention calculates earlier import the data each point square, utilize the method for recursion to calculate the interior energy of the corresponding sliding window of each reference point then, can avoid double counting to energy value

Description of drawings

Fig. 1 is basic principle figure of the present invention

Fig. 2 is a concrete enforcement block diagram of the present invention

Fig. 3 be do before and after the received signal relevant method and emulation comparison diagram of the present invention (signal to noise ratio :-3dB)

Fig. 4 is received signal and relevant method and the emulation comparison diagram (frequency deviation: 80Hz) of the present invention of local work

Fig. 5 is a simulation result figure of the present invention (signal to noise ratio :-3dB frequency deviation: 80Hz)

Specific implementation method

The present invention is verified at an ofdm system that is used for short wave communication, provides the parameter concrete configuration of system synchronization part below, and sets forth specific implementation step of the present invention.

The synchronous head parameter:

Sample rate: 40.96KHz

Signal bandwidth: 12KHz

Every section synchronizing sequence length: 256 points

Synchronizing sequence hop count: 8 sections

The receiving sequence segmentation is counted: 1792 points

At first the counter-rotating of 8 sections 256 synchronizing sequences is got conjugation then zero padding become be FFT again at 2048, be stored in the local reception end, store the quadratic sum of local each section of sequence simultaneously.

Synchronizing process is divided into following steps:

Step 1: receiving sequence is divided into one section according to 1792, every segment data again with the last period data 256 at end form 2048 point sequences (the 1st segment data front zero padding).

Step 2: 2048 point sequences after the segmentation are done to multiply each other with 8 sections local sequences respectively behind the FFT, again multiplied result is made IFFT respectively, cast out front 256 points, just obtain 8 sections 1792 correlation curve.

Step 3: calculate every of input data square, obtain the interior energy value of the corresponding sliding window of every spot correlation value with the method for recursion, obtain the energy value of correlation correspondence with the energy value of the local sequence evolution that multiplies each other respectively again.Correlation curve with the energy value normalization of trying to achieve, is obtained normalized correlation.

Step 4: with the storage of each section normalization correlation, delayed addition asks average then.Correlation time-delay 256* (8-1) point of the 1st section synchronizing sequence correspondence, correlation time-delay 256* (8-2) point of the 2nd section synchronizing sequence correspondence, and the like, just obtain final correlation curve after averaging.

Step 5: setting thresholding is 0.16, seeks peak point, finishes synchronously.

Provide simulation result below:

Fig. 5 is simulation result figure.Simulated conditions is signal to noise ratio 0dB, two footpaths, and time-delay 2ms, Doppler expands 1Hz, frequency deviation 80Hz.Simulation result shows, under the very poor situation of channel condition, still can obtain bigger relevant peaks.

Claims

1, a kind of method for synchronizing time of radio communication comprises the steps:

(1), synchronizing sequence is divided into N segment, making FFT after every segment data conjugation counter-rotating, and the energy value of the result of FFT and every segment data is stored in receiving terminal in advance,

(2), according to overlap-save method, will receive data sementation and make FFT, the section with section between have necessarily overlapping.Remake the correlation that IFFT just obtains receiving sequence and local N segment data after the result of FFT and local N segment data multiplied each other respectively,

(3), according to the energy of receiving sequence and the energy of local sequence, calculate the energy value of corresponding each correlation with the method for recursion.Use the correlation of calculating in the step (2) to obtain normalized correlation then divided by energy value,

(4), the N section correlation that the obtains addition again of delaying time is respectively asked on average, find out relevant peaks, find out the original position of signal with the method that peak value detects.

2, method for synchronizing time according to claim 1 is characterized in that local synchronizing sequence is divided into N isometric segment, calculates the correlation of receiving sequence and N segment, finds out their common peak values.

3, method for synchronizing time according to claim 1 is characterized in that a receiving sequence and N segment combines the thought of the overlap-save method of fast convolution when asking correlation, be converted into 1 FFT computing and N IFFT computing with the computing of N correlations.