CN1599367A

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

Info

Publication number: CN1599367A
Application number: CN 200410041478
Authority: CN
Inventors: 吴蒙; 朱琦; 邵世祥; 张艳; 酆广增; 赵夙
Original assignee: Nanjing Post & Telecommunication College
Current assignee: Nanjing Post & Telecommunication College
Priority date: 2004-07-26
Filing date: 2004-07-26
Publication date: 2005-03-23
Anticipated expiration: 2024-07-26
Also published as: CN100389582C

Abstract

The synchronization method in the Orthogonal Frequency Division Multiplexing broadband wireless access system relates to a kind of Orthogonal Frequency Division Multiplexing (OFDM) synchronization implementation scheme specially used for 802.16a broadband wireless access network, and the synchronization steps are: using frame The first one makes delay-related judgments to achieve rough synchronization; utilizes the conjugate symmetry of frame header two to achieve precise timing; uses frame header one to perform rough estimation of fractional frequency offset, and uses frame header two to perform fine estimation of fractional frequency offset; uses frame header 2 and the frequency domain sequence of the frame header 1 estimate the integer frequency offset; the last synchronous tracking; wherein: the frame header includes a long training sequence and a short training sequence, the frame header 1 is a short training sequence including 4 repeated cycles, and the frame header 2 The long training sequence includes 2 repetition periods, which are generated by inserting 0 before the IFFT transformation. The frame header 2 not only has periodic repetition, but also has conjugate symmetry.

Description

Orthogonal Frequency Division Multiplexing Synchronization Method in Broadband Wireless Access System

技术领域Technical field

本发明涉及一种特别用于802.16a宽带无线接入网的正交频分复用(OFDM)同步实现方案，属于通信技术领域。The invention relates to a synchronization realization scheme of Orthogonal Frequency Division Multiplexing (OFDM) especially for 802.16a broadband wireless access network, and belongs to the technical field of communication.

背景技术 Background technique

OFDM系统因其具有抗干扰能力强、频率利用率高等优点而越来越得到广泛关注，无线宽带接入系统806.16a就采用了OFDM技术作为系统的传输方式之一。但是它对符号同步要求非常高：OFDM使用正交的子载波提高频谱利用率，如果定时不同步会破坏子载波的正交性；频率偏移引起有用信号的相位旋转和幅度衰落，更严重的是造成子载波间干扰，系统性能将大幅度下降，从而补偿由于时间和频率不同步带来的系统性能下降非常重要的。The OFDM system has attracted more and more attention because of its advantages of strong anti-interference ability and high frequency utilization rate. The wireless broadband access system 806.16a adopts OFDM technology as one of the transmission methods of the system. However, it has very high requirements for symbol synchronization: OFDM uses orthogonal subcarriers to improve spectrum utilization, and if the timing is not synchronized, the orthogonality of subcarriers will be destroyed; frequency offset causes phase rotation and amplitude fading of useful signals, and more serious It is very important to compensate for the system performance degradation caused by time and frequency asynchronization due to the interference between sub-carriers and the system performance will be greatly reduced.

近几年来，人们对OFDM同步进行了广泛的研究，其中比较典型的算法主要包括以下一些：Moose给出了频域最大似然估计方法估计频偏，其实质是利用符号的重复进行频域估计，Moose这种基于重复相关的思想为后来的频偏估计算法提供了基本的思路，但由于他的算法要求帧头使用两个重复的符号且信道相干时间至少大于两个符号的持续时间，故在现在很多实际系统中并不用此方法；Nogami和Nagashima采用了一种null symbol，在接收端通过检测到功率下降来帧同步，但这方法适用于连续结构，对于突发数据结构，由于无法正确区分Null symbol和空闲时间而不宜采用此方法；Van de Beek提出了基于CP的最大似然估计算法，该算法无法估计整数频偏，且不能提供帧同步信息，因而多用于定时和频偏的跟踪阶段，考虑到同步跟踪要保证算法简单，有许多文献都在此基础上提出了各种改进算法；M.Schmidl提出了基于训练序列的方法，他利用帧头两个训练序列实现了能量检测和精同步以及频偏估计，该算法比较简单且易实现；Classen还提出了基于导频方法，其前提是慢时变信道，该算法在频域估计频偏时已经引入了ICI，故性能有所下降；还有目前正在研究的盲估计也较多，由于其基于多个符号的平均，要求信道慢时变，且算法复杂多用于频偏跟踪。In recent years, people have conducted extensive research on OFDM synchronization, among which the typical algorithms mainly include the following: Moose gives the frequency domain maximum likelihood estimation method to estimate the frequency offset, which essentially uses the repetition of symbols to estimate the frequency domain , Moose's idea based on repeated correlation provided the basic idea for the subsequent frequency offset estimation algorithm, but because his algorithm requires the frame header to use two repeated symbols and the channel coherence time is at least greater than the duration of two symbols, so This method is not used in many actual systems now; Nogami and Nagashima use a null symbol, which detects the power drop at the receiving end to synchronize the frame, but this method is suitable for continuous structures. For burst data structures, it cannot be correct This method is not suitable for distinguishing between Null symbol and idle time; Van de Beek proposed a maximum likelihood estimation algorithm based on CP, which cannot estimate integer frequency offset and cannot provide frame synchronization information, so it is mostly used for timing and frequency offset tracking stage, considering that the synchronous tracking should ensure a simple algorithm, many literatures have proposed various improved algorithms on this basis; M.Schmidl proposed a method based on training sequences, which realized energy detection and Fine synchronization and frequency offset estimation, the algorithm is relatively simple and easy to implement; Classen also proposed a pilot-based method, the premise of which is a slow time-varying channel. This algorithm has introduced ICI when estimating frequency offset in the frequency domain, so the performance is somewhat different. There are also many blind estimates currently being studied, because they are based on the average of multiple symbols, requiring slow time-varying channels, and the algorithm is complex and mostly used for frequency offset tracking.

发明内容Contents of Invention

技术问题：本发明的目的是提供一种宽带无线接入系统中正交频分复用的同步方法，该方法在系统设计上基于训练序列，充分利用帧头提供的信息，对精定时同步和整数频偏估计方法进行了改善，调整了同步顺序，从而降低了计算量，且提高了频偏估计性能。Technical problem: The purpose of the present invention is to provide a synchronization method for OFDM in a broadband wireless access system. The method is based on the training sequence in system design, fully utilizes the information provided by the frame header, and performs accurate timing synchronization and The integer frequency offset estimation method is improved, and the synchronization order is adjusted, thereby reducing the calculation amount and improving the frequency offset estimation performance.

技术方案：本发明的OFDM系统中的同步方法：利用帧头一做延迟相关判决，实现粗同步；利用帧头二的共扼对称性，实现精定时；用帧头一进行小数频偏粗估计，用帧头二进行小数频偏精估计；用帧头二与帧头一的频域序列估计整数频偏；最后同步跟踪。Technical solution: the synchronization method in the OFDM system of the present invention: use frame header 1 to make delay-related judgments to realize rough synchronization; utilize the conjugate symmetry of frame header 2 to realize precise timing; use frame header 1 to perform coarse estimation of fractional frequency offset , use frame header 2 to perform precise estimation of fractional frequency offset; use the frequency domain sequences of frame header 2 and frame header 1 to estimate integer frequency offset; and finally track synchronously.

其中：帧头包括了长训练序列和短训练序列，帧头1为短训练序列包括4个重复的周期，而帧头2为长训练序列包括2个重复周期，这种重复周期性通过在反快速傅立叶变换IFFT变换之前插0生成。Among them: the frame header includes the long training sequence and the short training sequence, the frame header 1 includes 4 repetition periods for the short training sequence, and the frame header 2 includes 2 repetition periods for the long training sequence. The fast Fourier transform IFFT is generated by inserting 0 before the transform.

粗同步的方法为：利用帧头一的周期性，采用延迟相关方法(即求一个信号与其经过时延以后信号的相关函数)，有两个滑动窗，一个是计算接收信号与其延迟信号的相关函数，延迟系数等于短训练符号的周期长度64；另一个计算相关窗内的信号能量，是用来对判决变量进行归一化的，这样判决变量就不再依赖于能量的绝对水平，当判决变量大于门限值时，可将该处确定为初同步处。The method of coarse synchronization is: use the periodicity of frame header 1, adopt the delay correlation method (that is, find the correlation function between a signal and its delayed signal), there are two sliding windows, one is to calculate the correlation between the received signal and its delayed signal function, the delay coefficient is equal to the period length of the short training symbol 64; the other calculates the signal energy in the correlation window, which is used to normalize the decision variable, so that the decision variable no longer depends on the absolute level of energy, when the decision When the variable is greater than the threshold value, this location can be determined as the initial synchronization location.

精定时的方法为：由于本发明的帧头二不但具有周期重复性，而且具有共扼对称性，帧头结构如[A/4 B/4 A/4 B/4]，而B关于A共扼对称。正是利用这种共扼对称性，求前一半帧头和后一半帧头的共扼相关函数作为判决变量，搜索该判决变量的最大值，这个最大值处即为精同步处，也就是一帧的精确起点。精定时算法比求互相关算法计算量本身就减少了一半的乘法，而且不因为搜索整数频偏而反复计算，计算量更是大大减少，另外一个好处就是利用共扼对称性求最大相关峰时不受频偏影响，提高了同步的精度。The method of precise timing is: because the frame header 2 of the present invention not only has periodic repeatability, but also has conjugate symmetry, the frame header structure is such as [A/4 B/4 A/4 B/4], and B is related to A Symmetry. Just using this conjugate symmetry, find the conjugate correlation function of the first half of the frame header and the second half of the frame header as the decision variable, and search for the maximum value of the decision variable. The maximum value is the fine synchronization point, that is, a The exact starting point of the frame. Compared with the cross-correlation algorithm, the calculation amount of the precise timing algorithm itself is reduced by half, and the calculation amount is greatly reduced without repeated calculations because of searching for integer frequency offsets. Another advantage is to use conjugate symmetry to find the maximum correlation peak time. It is not affected by frequency offset, which improves the synchronization accuracy.

小数频偏估计的方法为：用帧头一进行小数频偏粗估计，用帧头二进行小数频偏精估计。对于短训练序列而言，可估计的最大频偏为2；对于长训练序列，其D＝126，可估计的最大频偏为1。The method of estimating the fractional frequency offset is: use frame header 1 to perform rough estimation of the fractional frequency offset, and use frame header 2 to perform fine estimation of the fractional frequency offset. For the short training sequence, the maximum estimable frequency offset is 2; for the long training sequence, D=126, the maximum estimable frequency offset is 1.

整数频偏估计的方法为：采用帧头二与帧头一频域联合估计整数频偏。因为小数频偏的纠正避免了ICI，从而频域符号仅因为整数频偏产生移位，且帧头一与帧头二的移位一致，移动位数为4的倍数；该方法利用帧头二与帧头一频域4倍子载波上的比值形成新的已知PN序列v_k，与接收到的相应PN序列做相关，在可能的整数频偏范围内每次移位4个采样点搜索最大值，确定最大值时的移位数，即为整数频偏系数。The method for estimating the integer frequency offset is as follows: jointly estimating the integer frequency offset by using frame header 2 and frame header 1 frequency domain. Because the correction of the fractional frequency offset avoids ICI, the frequency domain symbols are shifted only because of the integer frequency offset, and the shift of frame header 1 and frame header 2 is consistent, and the number of shifting bits is a multiple of 4; this method uses frame header 2 Form a new known PN sequence v _k with the ratio of the frame head-frequency domain 4 times the subcarrier, correlate with the received corresponding PN sequence, shift 4 sampling points each time within the possible integer frequency offset range to search The maximum value, the number of shifts when the maximum value is determined, is the integer frequency offset coefficient.

同步跟踪的方法为：利用OFDM信号中所固有的循环前缀里信息冗余的特征，采用最大似然估计器在时域进行符号定时和频偏的联合估计。The method of synchronous tracking is as follows: using the feature of information redundancy in the cyclic prefix inherent in OFDM signals, the maximum likelihood estimator is used to jointly estimate symbol timing and frequency offset in the time domain.

有益效果：本发明基于802.16a系统的特点提出了一套整体的同步方法，不但降低了算法复杂度，而且改善了系统的性能，其中定时精同步算法利用了802.16a帧头的共扼对称性，比传统的精定时方法计算量小且对频偏不敏感，因而可以在频率同步之前完成，使各种同步之间克服了相互依赖性，提高了系统的鲁棒性。该方法在系统设计上基于训练序列，充分利用帧头提供的信息，对精定时同步和整数频偏估计方法进行了改善，调整了同步顺序，从而降低了计算量，且提高了频偏估计性能。Beneficial effects: the present invention proposes a set of overall synchronization methods based on the characteristics of the 802.16a system, which not only reduces the complexity of the algorithm, but also improves the performance of the system, wherein the timing fine synchronization algorithm utilizes the conjugate symmetry of the 802.16a frame header , which is less computationally intensive than the traditional precise timing method and is insensitive to frequency offset, so it can be completed before frequency synchronization, which overcomes the interdependence between various synchronizations and improves the robustness of the system. Based on the training sequence in system design, this method makes full use of the information provided by the frame header, improves the precise timing synchronization and integer frequency offset estimation method, adjusts the synchronization sequence, thereby reducing the amount of calculation and improving the performance of frequency offset estimation .

附图说明Description of drawings

图1是本发明帧头结构图。Fig. 1 is a frame header structure diagram of the present invention.

图2是本发明同步方案图。Fig. 2 is a diagram of the synchronization scheme of the present invention.

图3是本发明延迟相关示意图。Fig. 3 is a schematic diagram of delay correlation in the present invention.

图4是本发明粗同步框图。Fig. 4 is a block diagram of coarse synchronization in the present invention.

图5是本发明延迟相关判决变量响应图。Fig. 5 is a response diagram of delay-related decision variables in the present invention.

图6是本发明精同步框图。Fig. 6 is a fine synchronization block diagram of the present invention.

图7是本发明最大似然估计器结构图。Fig. 7 is a structural diagram of the maximum likelihood estimator of the present invention.

图8是本发明同步跟踪框图。Fig. 8 is a synchronous tracking block diagram of the present invention.

以上图中有：延迟相关滑动窗C，信号能量计算滑动窗P，c(n)是接收信号延迟相关函数，r(n)是接收到的信号，p(n)是接收信号能量函数，M(n)是时间同步函数，Z^-D为延迟系数，+为相加，×为相乘，÷为相除，||²为求绝对平方，()²为求平方，()^*为求共扼，θ_ML为符号定时，ε_ML为频偏估计，γ(θ)是接收信号N位延迟相关函数，Φ()为移位求和的结果，ρ||²/2为求绝对平方的一半，||为求绝对值，∠为求角度，-1/2π为信号乘的倍数，Arg max为求最大值，r(k)是接收到的信号。In the above figure, there are: delay correlation sliding window C, signal energy calculation sliding window P, c(n) is the delay correlation function of the received signal, r(n) is the received signal, p(n) is the energy function of the received signal, M (n) is the time synchronization function, Z ^-D is the delay coefficient, + is addition, × is multiplication, ÷ is division, || ² is for absolute square, () ² is for square, () ^* is for Conjugate, θ _ML is the symbol timing, ε _ML is the frequency offset estimation, γ(θ) is the N-bit delay correlation function of the received signal, Φ() is the result of shift summation, ρ|| ² /2 is the absolute square Half of , || is to find the absolute value, ∠ is to find the angle, -1/2π is the multiple of signal multiplication, Arg max is to find the maximum value, and r(k) is the received signal.

具体实施方式 Detailed ways

本发明帧头结构如图1所示，包括了长训练序列和短训练序列，其中帧头1为短训练序列包括4个重复的周期，每个周期有64个抽样点，而帧头2为长训练序列包括2个重复周期，每个周期有128个抽样点，其中CP为循环前缀。The frame header structure of the present invention as shown in Figure 1, has included long training sequence and short training sequence, and wherein frame header 1 is that short training sequence comprises 4 repeated cycles, and each cycle has 64 sampling points, and frame header 2 is The long training sequence includes 2 repetition periods, each period has 128 sampling points, where CP is the cyclic prefix.

具体的方案如图2，主要分为以下五个步骤：The specific plan is shown in Figure 2, which is mainly divided into the following five steps:

(1)粗同步(1) Rough synchronization

粗同步利用了帧头一的周期性，这种算法可称为延迟相关算法，如图3所示。Coarse synchronization utilizes the periodicity of frame header 1, and this algorithm can be called a delay-related algorithm, as shown in Figure 3.

图中显示有两个滑动窗C和P，C窗是接收信号与其延迟信号的相关函数，因此称为延迟相关，延迟系数Z-D在本发明中等于短训练符号的周期长度64；P窗进行相关窗内的信号能量计算，是用来对判决变量进行归一化的，这样判决变量就不再依赖于能量的绝对水平。Two sliding windows C and P are shown in the figure, and the C window is the correlation function of the received signal and its delayed signal, so it is called delay correlation, and the delay coefficient Z-D is equal to the period length 64 of the short training symbol in the present invention; the P window is correlated The signal energy calculation within the window is used to normalize the decision variable so that the decision variable is no longer dependent on the absolute level of energy.

设c(n)是接收信号延迟相关函数，r(n)是接收到的信号，p(n)是接收信号能量函数，M(n)是时间同步函数，L是窗的长度，一般取帧头重复序列中每个序列的长度，数据帧开始的粗估计位置n_coarse。Let c(n) be the delay correlation function of the received signal, r(n) be the received signal, p(n) be the energy function of the received signal, M(n) be the time synchronization function, L be the length of the window, generally take the frame The length of each sequence in the header repeat sequence, the coarse estimated position n _coarse of the start of the data frame.

$c c ((n no)) = = {Σ Σ}_{k k = = 00}^{L L - - 11} r r ((n no + + k k)) r r * * ((n no + + k k + + D D.)) - - - - - - ((11))$

$p p ((n no)) = = {Σ Σ}_{k k = = 00}^{L L - - 11} {| | r r ((n no + + k k + + D D.)) | |}^{22} - - - - - - ((22))$

时间同步函数定义为The time synchronization function is defined as

$M m ((n no)) = = \frac{{| | c c ((n no)) | |}^{22}}{{((p p ((n no))))}^{22}} - - - - - - ((33))$

由于c为延迟相关滑动窗，p为计算信号能量滑动窗，迭代的办法可以用来减少计算量。Since c is a delay-related sliding window and p is a sliding window for calculating signal energy, an iterative method can be used to reduce the amount of calculation.

具体如图4所示：先利用接收到的数据r(n)的前64个数据点计算c(n)，p(n)和M(n)值，即 $c (1) = Σ_{k = 0}^{63} r (1 + k) r^{*} (1 + k + 64),$ $p (1) = Σ_{k = 0}^{63} {| r (1 + k + 64) |}^{2}$ 和Specifically as shown in Figure 4: first use the first 64 data points of the received data r(n) to calculate c(n), p(n) and M(n) values, that is $c (1) = Σ_{k = 0}^{63} r (1 + k) r^{*} (1 + k + 64),$ $p (1) = Σ_{k = 0}^{63} {| r (1 + k + 64) |}^{2}$ and

$M m ((11)) = = \frac{{| | c c ((11)) | |}^{22}}{{((p p ((11))))}^{22}};;$

然后进行迭代：Then iterate:

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

$M m ((22)) = = \frac{{| | c c ((22)) | |}^{22}}{{((p p ((22))))}^{22}}$

以此类推，可以计算一组(假设1000个)M(n)值。图5为本发明的前导序列在SNR＝10dB环境下延迟相关判决变量的响应，其中横轴为n，纵轴为M(n)。由图可见当数据帧开始时，M(n)迅速跳变到最大值，这个跳变可极为有效的用来判定数据帧开始的粗估计位置n_coarse。该方法受短训练符号序列自身的统计特性和噪声影响，而与频偏无关，所以在频率同步之前进行粗同步。实验证明在SNR＞6dB的情况下，系统粗同步性能比较理想。By analogy, a set (assuming 1000) of M(n) values can be calculated. Fig. 5 is the response of the preamble sequence of the present invention to the delay-related decision variable under the environment of SNR=10dB, wherein the horizontal axis is n, and the vertical axis is M(n). It can be seen from the figure that when the data frame starts, M(n) quickly jumps to the maximum value, and this jump can be used to determine the rough estimated position n _coarse of the start of the data frame very effectively. This method is affected by the statistical characteristics and noise of the short training symbol sequence itself, but has nothing to do with the frequency offset, so coarse synchronization is performed before frequency synchronization. The experiment proves that in the case of SNR>6dB, the coarse synchronization performance of the system is ideal.

(2)精同步(2) Fine synchronization

本发明的帧头二不但具有周期重复性，而且具有共扼对称性，帧头结构如[A/4 B/4 A/4 B/4]，而B关于A共扼对称。正是利用这种共扼对称性，精定时算法作如下改进：The frame header 2 of the present invention not only has periodic repetition, but also has conjugate symmetry. The frame header structure is such as [A/4 B/4 A/4 B/4], and B is conjugate symmetric about A. It is by using this conjugate symmetry that the precise timing algorithm is improved as follows:

设P(d)是接收信号自相关函数，R(d)是接收信号能量函数，r(n)是接收到的信号。N是一个OFDM符号里的数据的个数，M_Pro(d)是精同步算法的时间同步函数。Let P(d) be the received signal autocorrelation function, R(d) be the received signal energy function, and r(n) be the received signal. N is the number of data in one OFDM symbol, and M _Pro (d) is the time synchronization function of the fine synchronization algorithm.

${M m}_{Pro Pro} ((d d)) = = \frac{{| | P P ((d d)) | |}^{22}}{{((R R ((d d))))}^{22}} - - - - - - ((44))$

其中， $P (d) = Σ_{k = 0}^{N / 4 - 1} r (d - k) r (d + k) - - - (5)$ in, $P (d) = Σ_{k = 0}^{N / 4 - 1} r (d - k) r (d + k) - - - (5)$

$R R ((d d)) = = {Σ Σ}_{k k = = 00}^{N N / / 44 - - 11} {| | r r ((d d + + k k)) | |}^{22} - - - - - - ((66))$

此时精定时算法比求互相关算法计算量本身就减少了一半的乘法，而且不因为搜索整数频偏而反复计算，计算量更是大大减少。At this time, the calculation amount of the precise timing algorithm is reduced by half compared with the multiplication of the cross-correlation algorithm itself, and the calculation amount is greatly reduced without repeated calculations because of searching for the integer frequency offset.

具体步骤如图6所示：根据粗同步得到的粗估计同步位置n_coarse，得到帧头二CP后的第一位数据的位置是(n_coarse+32+256+32)＝n_coarse+320，而帧头二是以第65位数据即n_coarse+320+64＝n_coarse+384处为中心对称，粗同步中也提到估计到的位置偏早，于是计算 $R (n_{coarse} + 384) = Σ_{k = 0}^{N / 4 - 1} {| r (n_{coarse} + 384 + k) |}^{2}$ 和The specific steps are as shown in Figure 6: according to the coarse estimated synchronization position n _coarse obtained by the coarse synchronization, the position of the first bit of data after the frame header 2 CP is obtained is (n _coarse +32+256+32)=n _coarse +320, The frame header 2 is symmetrical to the center of the 65th bit data, that is, n _coarse +320+64=n _coarse +384. It is also mentioned in the coarse synchronization that the estimated position is too early, so the calculation $R ({no}_{coarse} + 384) = Σ_{k = 0}^{N / 4 - 1} {| r ({no}_{coarse} + 384 + k) |}^{2}$ and

$P P (({n no}_{coarse coarse} + + 384384)) = = {Σ Σ}_{k k = = 00}^{N N / / 44 - - 11} (({n no}_{coarse coarse} + + 384384 - - k k)) r r (({n no}_{coarse coarse} + + 384384 + + k k))$

得到： $M_{Pro} (n_{coarse} + 384) = \frac{{| P (n_{coarse} + 384) |}^{2}}{{(R (n_{coarse} + 384))}^{2}},$ 然后计算get: $m_{Pro} ({no}_{coarse} + 384) = \frac{{| P ({no}_{coarse} + 384) |}^{2}}{{(R ({no}_{coarse} + 384))}^{2}},$ then calculate

$P P (({n no}_{coarse coarse} + + 385385)) = = {Σ Σ}_{k k = = 00}^{N N / / 44 - - 11} r r (({n no}_{coarse coarse} + + 385385 - - k k)) r r (({n no}_{coarse coarse} + + 385385 + + k k))$

得到： $M_{Pro} (n_{coarse} + 385) = \frac{{| P (n_{coarse} + 385) |}^{2}}{{(R (n_{coarse} + 385))}^{2}}$ get: $m_{Pro} ({no}_{coarse} + 385) = \frac{{| P ({no}_{coarse} + 385) |}^{2}}{{(R ({no}_{coarse} + 385))}^{2}}$

以此类推，得到一组(比如32个)M_Pro的值，其中最大值所对应的序号n_n减64等于n_fine2也就是精同步所要找的同步位置n_head2，也就是帧头二的CP后的第一位数据的位置，则帧头一CP后的第一位数据的位置n_head1是n_fine2-256-32＝n_fine2-288。By analogy, a group (such as 32) of M _Pro values is obtained, and the sequence number corresponding to the maximum value _n minus 64 is equal to n _fine2 , which is the synchronization position n _head2 that fine synchronization is looking for, that is, the CP of frame head 2 The position of the first bit of data after CP, then the position n _head1 of the first bit of data after the frame header CP is n _fine2 -256-32=n _fine2 -288.

此方法的另外一个好处就是利用共扼对称性求最大相关峰时不受频偏影响，设发送信号为x(n)，忽略掉噪声的影响，接受的基带信号r(n)＝x(n)e^j2πnε/N。由(5)式可推得：Another advantage of this method is that it is not affected by frequency offset when using conjugate symmetry to find the maximum correlation peak. Let the transmitted signal be x(n), ignore the influence of noise, and receive the baseband signal r(n)=x(n )e ^j2πnε/N . From formula (5), it can be deduced that:

${| | P P ((d d)) | |}^{22} = = {| | {Σ Σ}_{k k = = 00}^{N N / / 44 - - 11} r r ((d d - - k k)) r r ((d d + + k k)) | |}^{22} = = {| | {Σ Σ}_{k k = = 00}^{N N / / 44 - - 11} x x ((d d - - k k)) {e e}^{\frac{j j 22 πϵ πϵ ((d d - - k k))}{256256}} x x ((d d + + k k)) {e e}^{\frac{j j 22 πϵ πϵ ((d d + + k k))}{256256}} | |}^{22}$

上式可见，最大相关峰值函数与频偏无关，因此精定时同步可以放在频偏估计之前，不但解决了小数频偏估计均方误差大的问题，而且避免了定时与频偏错误相互影响，使各种同步算法更加独立，大大了提高系统的鲁棒性。It can be seen from the above formula that the maximum correlation peak function has nothing to do with frequency offset, so precise timing synchronization can be placed before frequency offset estimation, which not only solves the problem of large mean square error in fractional frequency offset estimation, but also avoids the mutual influence of timing and frequency offset errors. Make various synchronization algorithms more independent, greatly improving the robustness of the system.

(3)小数频偏估计(3) Fractional frequency offset estimation

设收发端的绝对频率偏差为Δf，系统抽样率为f_s，N是子载波数，相对于子载波间隔的频偏系数定义为 $ϵ = \frac{Δf}{f_{s} / N},$ D帧头的重复序列中每个序列的长度，z为延迟相关滑动函数，帧头一CP后的第一位数据的位置n_head1，帧头二CP后的第一位数据的位置n_head2，小数粗频偏为ε_coarse，小数精频偏为ε_fine。Assuming the absolute frequency deviation of the transceiver end is Δf, the system sampling rate is f _s , N is the number of subcarriers, and the frequency offset coefficient relative to the subcarrier spacing is defined as $ϵ = \frac{Δf}{f_{the s} / N},$ The length of each sequence in the repeated sequence of the D frame header, z is the delay-related sliding function, the position n _head1 of the first bit of data after the first CP of the frame header, and the position n _head2 of the first bit of data after the second CP of the frame header, The decimal coarse frequency offset is ε _coarse , and the decimal fine frequency offset is ε _fine .

本发明采用的小数频偏估计器为The fractional frequency offset estimator adopted in the present invention is

$ϵ ϵ = = - - \frac{N N}{22 πD πD} angle the angle ((z z)) - - - - - - ((88))$

设发送信号为x(n)，忽略掉噪声的影响，接受的基带信号r(n)＝x(n)e^j2πnε/N。Let the transmitted signal be x(n), ignore the influence of noise, and receive the baseband signal r(n)=x(n)e ^j2πnε/N .

$z z = = {Σ Σ}_{n no = = 00}^{D D. - - 11} r r ((n no)) {r r}^{* *} ((n no + + D D.))$

$= = {Σ Σ}_{n no = = 00}^{D D. - - 11} x x ((n no)) {x x}^{* *} ((n no + + D D.)) {e e}^{j j 22 πϵn πϵn / / N N} {e e}^{- - j j 22 πϵ πϵ ((n no + + D D.)) / / N N}$

$= = {e e}^{- - j j 22 πϵD πϵD / / N N} {Σ Σ}_{n no = = 00}^{D D. - - 11} {| | x x ((n no)) | |}^{22} - - - - - - ((99))$

由于angle(z)是定义在区间[-ππ]上的，因此可估计的频偏范围为Since angle(z) is defined on the interval [-ππ], the estimated frequency offset range is

$| | ϵ ϵ | | \leq \leq \frac{N N}{22 D D.} - - - - - - ((1010))$

对于短训练序列而言，其D＝64，因此可估计的最大频偏为2；对于长训练序列，其D＝126，可估计的最大频偏为1。For the short training sequence, D=64, so the maximum estimable frequency offset is 2; for the long training sequence, D=126, the maximum estimable frequency offset is 1.

由于在粗同步中已经计算过 $c (n) = Σ_{k = 0}^{63} r (n + k) r^{*} (n + k + 64),$ 其中n＝1...1000，则由帧头一做小数频偏粗估计时，c(n)不需要再计算，直接引用粗同步中已经计算出的值来估计小数频偏，则小数粗频偏为：Since it has been calculated in the coarse synchronization $c (no) = Σ_{k = 0}^{63} r (no + k) r^{*} (no + k + 64),$ Where n=1...1000, when the fractional frequency offset is roughly estimated by frame header 1, c(n) does not need to be calculated again, and the value already calculated in the coarse synchronization is directly used to estimate the fractional frequency offset, then the fractional frequency offset is roughly The frequency offset is:

${ϵ ϵ}_{coarse coarse} = = - - \frac{256256}{22 π π * * 6464} angle the angle ((c c (({n no}_{head the head 11}))))$

然后校正所有接收数据。校正后，就由帧头二做小数频偏精估计， $z = Σ_{k = 0}^{127} r (n_{head 2} + k) r^{*} (n_{head 2} + k + 128),$ 则小数精频偏为：All received data is then corrected. After correction, the fractional frequency offset is finely estimated by frame header 2, $z = Σ_{k = 0}^{127} r ({no}_{thehead 2} + k) r^{*} ({no}_{thehead 2} + k + 128),$ Then the fractional precision frequency deviation is:

${ϵ ϵ}_{fine fine} = = - - \frac{256256}{22 π π * * 128128} angle the angle ((z z))$

然后再校正一次数据。Then correct the data again.

(4)整数频偏估计(4) Integer frequency offset estimation

802.16a系统的载频范围为2.4G~11GHz，标准允许的最大晶振误差为20ppm，如果发射机和接收机的时钟都是在最大误差状态，且正负相反的话，则发射机和接收机之间的误差为40ppm，所以可能的最大频偏范围为：The carrier frequency range of the 802.16a system is 2.4G~11GHz, and the maximum crystal oscillator error allowed by the standard is 20ppm. If the clocks of the transmitter and receiver are in the state of maximum error, and the positive and negative are opposite, the difference between the transmitter and the receiver The error between is 40ppm, so the possible maximum frequency deviation range is:

ΔF＝40*10^-6*(2.4~11)*10⁹＝96~440KHzΔF＝40*10 ^-6 *(2.4~11)*10 ⁹ ＝96~440KHz

而子载波间隔为：And the subcarrier spacing is:

Δf＝(1.75*m*10⁶)/256＝6.8359*m KHzΔf＝(1.75*m*10 ⁶ )/256＝6.8359*m KHz

所以最大频偏系数可达65左右，仅仅是小数频偏纠正不够，因此还要进行整数频偏纠正。Therefore, the maximum frequency offset coefficient can reach about 65, only fractional frequency offset correction is not enough, so integer frequency offset correction is also required.

本发明采用帧头二与帧头一频域联合估计整数频偏。因为小数频偏的纠正避免了ICI，从而频域符号仅因为整数频偏产生移位，且帧头一与帧头二的移位一致，移动位数为4的倍数；该方法利用帧头二与帧头一频域4倍子载波上的比值形成新的已知PN序列v_k，与接收到的相应PN序列做相关，在可能的整数频偏范围内每次移位4个采样点搜索最大值，确定最大值时的移位数，即为整数频偏系数。设一PN序X＝{-W，-W+4，...，-4，0，4，...，W-4，W}，W是这个PN序列满足下标为4的倍数的个数，Y_1，n代表帧头一频域的第n个数据，Y_2，n代表帧头二频域的第n个数据，v_k为利用帧头二与帧头一频域4倍子载波上的比值形成新的已知PN序列。整数频偏估计函数为B(g)：The present invention adopts frame header two and frame header one frequency domain to jointly estimate integer frequency offset. Because the correction of the fractional frequency offset avoids ICI, the frequency domain symbols are shifted only because of the integer frequency offset, and the shift of frame header 1 and frame header 2 is consistent, and the number of shifting bits is a multiple of 4; this method uses frame header 2 Form a new known PN sequence v _k with the ratio of the frame head-frequency domain 4 times the subcarrier, correlate with the received corresponding PN sequence, shift 4 sampling points each time within the possible integer frequency offset range to search The maximum value, the number of shifts when the maximum value is determined, is the integer frequency offset coefficient. Suppose a PN sequence X={-W, -W+4, ..., -4, 0, 4, ..., W-4, W}, W is the PN sequence that satisfies the subscript as a multiple of 4 The number, Y _{1, n} represents the nth data in the frequency domain of frame header 1, Y _{2, n} represents the nth data in the frequency domain of frame header 2, and v _k is 4 times the frequency domain of frame header 2 and frame header 1 The ratios on the subcarriers form a new known PN sequence. The integer frequency offset estimation function is B(g):

$B B ((g g)) = = \frac{{| | {Σ Σ}_{k k &Element; &Element; X x} {Y Y}_{11,, k k + + 44 g g}^{* *} {v v}_{k k}^{* *} {Y Y}_{22,, k k + + 44 g g} | |}^{22}}{22 {(({Σ Σ}_{k k &Element; &Element; X x} {| | {Y Y}_{22,, k k} | |}^{22}))}^{22}} - - - - - - ((1111))$

具体步骤如下：先根据计算同步得到的同步位置，对接收信号的帧头做fft变换，得到帧头的频域信号Y_1，n和Y_2，n，然后计算 $v_{k} = \frac{Y_{1, k}}{Y_{2, k}} (k &Element; X) .$ 然后在适当的范围，比如-10~10(也就是搜寻的频偏范围时是-40~40)，计算 $B (g) = \frac{{| Σ_{k &Element; X} Y_{1, k + 4 g}^{*} v_{k}^{*} Y_{2, k + 4 g} |}^{2}}{2 {(Σ_{k &Element; X} {| Y_{2, k} |}^{2})}^{2}},$ 取其中B(g)最大者所对应的序号g，再乘以4，即为整数频偏。The specific steps are as follows: First, according to the synchronization position obtained by calculating the synchronization, perform fft transformation on the frame header of the received signal to obtain the frequency domain signals Y _1,n and Y _{2,n of the} frame header, and then calculate $v_{k} = \frac{Y_{1, k}}{Y} (k &Element; x) .$ Then in an appropriate range, such as -10~10 (that is, the frequency offset range of the search is -40~40), calculate $B (g) = \frac{{| Σ_{k &Element; x} Y_{1, k + 4 g}^{*} v_{k}^{*} Y_{2, k + 4 g} |}^{2}}{2 {(Σ_{k &Element; x} {| Y_{2, k} |}^{2})}^{2}},$ Take the sequence number g corresponding to the largest B(g) and multiply it by 4 to get the integer frequency offset.

(5)同步跟踪(5) Synchronous Tracking

OFDM符号为了对抗多径效应，引入了循环前缀作为保护间隔。循环前缀里的样值和符号末尾等长的一段样值是相同的，这种OFDM信号所固有的信息冗余特征为是实现符号定时和频率同步提供了可能。本发明采用最大似然估计器在时域进行对符号定时和频偏的联合估计进行跟踪，最大似然估计器结构如图7所示。In order to combat multipath effects, OFDM symbols introduce a cyclic prefix as a guard interval. The sample value in the cyclic prefix is the same as a sample value of equal length at the end of the symbol. The inherent information redundancy feature of this OFDM signal provides the possibility to realize symbol timing and frequency synchronization. The present invention uses a maximum likelihood estimator to track the joint estimation of symbol timing and frequency offset in the time domain. The structure of the maximum likelihood estimator is shown in FIG. 7 .

设N是一个OFDM符号里的数据的个数，γ(θ)是接收信号N位延迟相关函数，ζ(θ)是接收信号能量函数，L_cp是循环前缀CP的长度，SNR是信噪比， $ρ = \frac{SNR}{SNR + 1},$ 帧头一CP后的第一位数据的位置n_head1，帧头二CP后的第一位数据的位置n_head2，另外为了叙述方便定义变量α(θ)＝|γ(θ)|-ρζ(θ)。跟踪阶段的符号定时θ_ML和频偏估计ε_ML为：Suppose N is the number of data in an OFDM symbol, γ(θ) is the N-bit delay correlation function of the received signal, ζ(θ) is the energy function of the received signal, L _cp is the length of the cyclic prefix CP, and SNR is the signal-to-noise ratio , $ρ = \frac{SNR}{SNR + 1},$ The position n _head1 of the first bit of data after the first CP of the frame header, the position n _head2 of the first bit of data after the second CP of the frame header, and the variable α(θ)=|γ(θ)|-ρζ( θ). The symbol timing θ _ML and frequency offset estimation ε _ML in the tracking phase are:

${θ θ}_{ML ML} = = arg arg \underset{θ θ}{max max} ((| | γ γ ((θ θ)) | | - - ρζ ρζ ((θ θ)))) - - - - - - ((1212))$

${ϵ ϵ}_{ML ML} = = - - \frac{11}{22 π π} &angle; &angle; γ γ (({θ θ}_{ML ML})) - - - - - - ((1313))$

其中， $γ (θ) = Σ_{k = θ}^{θ + L_{cp} - 1} r (k) r^{*} (k + N) - - - (14)$ in, $γ (θ) = Σ_{k = θ}^{θ + L_{cp} - 1} r (k) r^{*} (k + N) - - - (14)$

$ζ ζ ((θ θ)) = = \frac{11}{22} {Σ Σ}_{k k = = θ θ}^{θ θ + + {L L}_{cp cp} - - 11} (({| | r r ((k k)) | |}^{22} + + {| | r r ((k k + + N N)) | |}^{22})) - - - - - - ((1515))$

定时同步算法不受频偏影响，且无需传输专门的同步符号，传输效率比较高；算法的计算量要求较小，易于实现。The timing synchronization algorithm is not affected by frequency offset, and does not need to transmit special synchronization symbols, so the transmission efficiency is relatively high; the algorithm requires less calculation and is easy to implement.

具体步骤如图8所示：根据同步得到的同步位置，找到帧头后第一位数据的序号n_data＝n_head1+288*2，因为实际的帧头后第一位数据的序号可能比n_data大，也可能比帧头后第一位数据的序号小，于是在n_data左右寻找更准确的同步位置，比如在n_data±32的范围内，于是先计算The specific steps are shown in Figure 8: According to the synchronization position obtained by synchronization, find the serial number of the first data after the frame header n _data = n _head1 +288*2, because the actual serial number of the first data after the frame header may be higher than n _{The data} is large, and it may be smaller than the serial number of the first data after the frame header, so look for a more accurate synchronization position around n _data , for example, within the range of n _data ±32, so calculate first

$γ γ ((- - 3232)) = = {Σ Σ}_{k k = = - - 3232}^{- - 3232 + + {L L}_{cp cp} - - 11} r r (({n no}_{data data} + + k k)) {r r}^{* *} (({n no}_{data data} + + k k + + 256256))$

$ζ ζ ((- - 3232)) = = \frac{11}{22} {Σ Σ}_{k k = = - - 3232}^{- - 3232 + + {L L}_{cp cp} - - 11} (({| | r r (({n no}_{data data} + + k k)) | |}^{22} + + {| | r r (({n no}_{data data} + + k k + + 256256)) | |}^{22}))$

α(-32)＝|γ(-32)|-ρζ(-32)α(-32)=|γ(-32)|-ρζ(-32)

然后根据粗同步中提到的迭代公式计算Then calculate according to the iterative formula mentioned in Rough Synchronization

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

$ζ ζ ((- - 3131)) = = ζ ζ ((- - 3232)) - - \frac{11}{22} (({| | r r (({n no}_{data data} - - 3232)) | |}^{22} + + {| | r r (({n no}_{data data} - - 3232 + + 256256)) | |}^{22}))$

$+ + \frac{11}{22} (({| | r r (({n no}_{data data})) | |}^{22} + + {| | r r (({n no}_{data data} + + 256256)) | |}^{22}))$

α(-31)＝|γ(-31)|-σζ(-31)α(-31)=|γ(-31)|-σζ(-31)

以此类推，得到一组α(θ)的值，其中最大值对应的序号θ，即为一个数据符号开始的位置θ_ML。则 $ϵ_{ML} = - \frac{1}{2 π} &angle; γ (θ_{ML}),$ 然后对此符号内所有位数据进行频偏补偿。同理在帧头后的第二个符号第三个符号等等，做同样的同步跟踪。By analogy, a set of α(θ) values is obtained, and the serial number θ corresponding to the maximum value is the position θ _ML at which a data symbol starts. but $ϵ_{ML} = - \frac{1}{2 π} &angle; γ (θ_{ML}),$ Then perform frequency offset compensation for all bit data in this symbol. Similarly, do the same synchronous tracking for the second symbol, the third symbol, etc. after the frame header.

Claims

1, a kind of synchronization method in OFDM broadband wireless access system, it is characterized in that the step of synchronization is:

a. Use frame header 1 to make delay-related judgments to achieve rough synchronization;

b. Utilize the conjugate symmetry of frame header 2 to realize precise timing;

c. Use frame header 1 to perform rough estimation of decimal frequency offset, and use frame header 2 to perform fine estimation of decimal frequency offset;

d. Estimate the integer frequency offset by using the frequency domain sequence of frame header 2 and frame header 1;

e. Final synchronous tracking;

Among them: the frame header includes a long training sequence and a short training sequence. The frame header 1 is a short training sequence including 4 repetition periods, while the frame header 2 is a long training sequence including 2 repetition periods. This repetition cycle is passed through the IFFT Generated by inserting 0 before the transformation, frame header 2 not only has periodic repetition, but also has conjugate symmetry.

2. The synchronization method in the OFDM broadband wireless access system according to claim 1 is characterized in that the coarse synchronization method is: utilize the periodicity of frame header 1 and adopt the delay correlation method, there are two Sliding window: delay correlation sliding window C and signal energy calculation sliding window P, one is to calculate the correlation function between the received signal and its delayed signal, the delay coefficient is equal to the period length of the short training symbol 64; the other is to calculate the signal energy in the correlation window, which is Used to normalize the decision variable so that the decision variable is no longer dependent on the absolute level of energy.

3. The synchronization method in the OFDM broadband wireless access system according to claim 1 is characterized in that the precise timing method is: utilize the conjugate symmetry of the frame header two to find the first half of the frame header and The conjugate correlation function of the second half of the frame header is used as a decision variable, and the maximum value of the decision variable is searched. The maximum value is the fine synchronization point, that is, the precise starting point of a frame.

4. The synchronization method in the OFDM broadband wireless access system according to claim 1, characterized in that the method for estimating the fractional frequency offset is: use frame header 1 to perform rough estimation of the fractional frequency offset, and use frame header Second, perform a precise estimation of the fractional frequency offset. For the short training sequence, the maximum estimable frequency offset is 2; for the long training sequence, the time delay D=128, the maximum estimable frequency offset is 1.

5. The synchronization method in the OFDM broadband wireless access system according to claim 1, characterized in that the integer frequency offset estimation method is as follows: using frame header two and frame header one frequency domain to jointly estimate the integer frequency offset In this method, a new known pseudo-random noise sequence PN sequence v _k is formed by using the ratio of frame header 2 and frame header 1 frequency domain 4 times subcarriers, and correlating with the received corresponding PN sequence, in the possible integer frequency Shift 4 sampling points each time in the offset range to search for the maximum value, and determine the shift number when the maximum value is determined, which is the integer frequency offset coefficient.

6. The synchronization method in the Orthogonal Frequency Division Multiplexing Broadband Wireless Access System according to claim 1, characterized in that the synchronization tracking method is: utilizing the feature of information redundancy in the inherent cyclic prefix in the OFDM signal, A maximum likelihood estimator is used to jointly estimate symbol timing and frequency offset in time domain.

7. The synchronization method in the OFDM broadband wireless access system according to claim 1, characterized in that the delay correlation judgment method is as follows: find a signal and the normalized correlation function of the signal after time delay as Decision variables, there are two sliding windows, one is the delay correlation sliding window C, which calculates the correlation function between the received signal and its delayed signal, and the delay coefficient is equal to the cycle length of the short training symbol 64; the other is the calculation of the signal energy sliding window P, which calculates the correlation The signal energy in the window is used to normalize the decision variable, so that the decision variable is no longer dependent on the absolute level of energy; when the decision variable is greater than the threshold value, it can be determined as the initial synchronization point.