WO2011063724A1

WO2011063724A1 - Synchronization searching method

Info

Publication number: WO2011063724A1
Application number: PCT/CN2010/078772
Authority: WO
Inventors: 许文; 黄学民
Original assignee: 苏州恩巨网络有限公司
Priority date: 2009-11-25
Filing date: 2010-11-16
Publication date: 2011-06-03
Also published as: CN101741800A; CN101741800B

Abstract

Provided in the present invention is a synchronization searching method, which includes: calculating a hysteresis autocorrelation function of data sample points of a received signal; according to the hysteresis autocorrelation function, determining a time offset measurement of the data sample points; according to the time offset measurement, calculating estimated values and of a time offset and a frequency offset of the data sample points; according to the estimated value or, performing cross correlation using the received signal and a primary synchronization channel P-SCH, so as to acquire the beginning of a P-SCH symbol and the beginning of half-frame.

Description

Synchronous search method

Technical field

The present invention relates to the field of communications, and in particular to a synchronous search method.

Background technique

In a wireless cellular system, a User Equipment (UE) will attempt to connect to the network, where synchronization is the first task. It is well known that OFDM (Orthogonal Frequency-Division Multiplexing) systems are very sensitive to frequency offsets and time offsets. Here, frequency offset and time offset refer to the frequency and time deviation between the received signal and the local reference signal used for signal demodulation. The frequency offset and time offset may be caused by the oscillator mismatch between the transmitter and the receiver, or by the Doppler effect, multipath propagation, and the like. Frequency offset can destroy the orthogonality between subcarriers and generate inter-carrier interference (ICI) and multiple access interference between carriers. Time offset can lead to severe inter-block interference (IBI). ). In order to avoid serious damage to the performance of the receiver, the frequency and time deviation must be “deterministically determined and fully compensated.” This is the main task of synchronization. It is one of the key technologies for implementing OFDM systems.

In an OFDM-based LTE (Long Term Evolution) system, the entire synchronization process is first DL (downlink) synchronization and then UL (Up Link) synchronization.

Downlink synchronization: In order to facilitate the synchronization of the terminal UE to the network, the base station eNB (eNode-B) periodically transmits a SCH (Synchronization Channel) signal and a PBCH (Physical Broadcast Channel) signal. The UE will estimate the initial time and frequency offset by frequency sweeping and detecting (usually) the strongest SCH signal. When this is successful, the UE can read some of the most basic system information such as Cell ID (Identity), system bandwidth, etc. in SCH and PBCH. In LTE, part of the basic information, the so-called MIB (Master information block), is transmitted via PBCH. The information included in the MIB includes the downlink bandwidth of the cellular, the structure of the PHICH (Hybrid Automatic Repeat Request), the physical HARQ indicator channel, and the SFN (System Frame Number). This information is necessary for the UE to complete the connection.

Uplink synchronization: The terminal UE transmits a time information that has been obtained in the downlink synchronization, that is, a so-called PRACH (Physical Random Access Channel). signal. The base station eNB calculates the transmission time of the UE according to the received PRACH signal. Then let the UE tamper with the transmission time and identify the identity of the UE, thereby completing the coarse synchronization.

In the process of implementing the present invention, the inventors have found that the synchronous search scheme in the prior art has higher complexity and lower efficiency.

Summary of the invention

The present invention aims to provide a synchronous search method, which can solve the problem of high complexity and low efficiency of the synchronous search scheme in the prior art.

In an embodiment of the present invention, a synchronous search method is provided, including the following steps: calculating a delayed autocorrelation function of a data sample point of a received signal;

Determining a time-biased metric of the data sample points according to the lag autocorrelation function;

Calculating the time offset of the data sample point and the estimated value of the frequency offset ε according to the time offset metric; and performing autocorrelation of the received signal with the primary synchronization channel P-SCH according to the estimated value, and obtaining the starting point and the field of the P-SCH symbol The starting point.

In the above embodiment, on the basis of the lag autocorrelation, the time-off metric of the data sample point is determined, and then the estimated value of the time-division and the frequency offset ε of the data sample point is calculated according to the time-biased metric, and according to the estimated value And auto-correlation of the received signal with the primary synchronization channel P-SCH, obtaining the starting point of the P-SCH symbol and the starting point of the field, reducing the complexity of the synchronous search, improving the efficiency, and overcoming the synchronous search scheme in the prior art. A problem of higher complexity and lower efficiency.

DRAWINGS

The drawings are intended to provide a further understanding of the present invention, and are intended to be illustrative of the invention. In the drawing:

1 shows a block diagram of an OFDM receiver with two receive (Rx) antennas in accordance with one embodiment of the present invention;

2 shows a frame structure diagram of a 3GPP LTE FDD according to an embodiment of the present invention;

3 is a flow chart showing a synchronous search method according to an embodiment of the present invention; FIG. 4 is a schematic diagram showing an overlay of metrics of a conventional CP in LTE according to an embodiment of the present invention;

FIG. 5 is a schematic diagram showing a P-SCH cross correlation method according to the related art; 6 shows a schematic diagram of a synchronization method detecting a starting point of a field according to an embodiment of the present invention;

Fig. 7 is a diagram showing the performance comparison of the synchronization method based on the P-SCH cross-correlation and the embodiment of the present invention according to the embodiments of Figs. 5 and 6.

Detailed ways

The invention will be described in detail below with reference to the drawings in conjunction with the embodiments.

1 shows a block diagram of an OFDM receiver with two receive (Rx) antennas in accordance with one embodiment of the present invention. The so-called Begin of Frame (BOF) and Carrier Frequency Offset (CFO) can be calculated by performing coarse synchronization. At the same time, the Beginning of OFDM (Bore of Symbol, BOS) needs to be determined. Here, unless otherwise indicated, the OFDM symbol includes a CP (Cyclic Prefix) portion in the time domain. Therefore BOS is also the starting point of CP. After determining the BOS, the CP can be properly deleted. Through DFT (Discrete Fourier Transformation), the signal is converted from the time domain to the frequency domain. In this way, channel estimation and equalization can be easily implemented in the frequency domain.

We consider an OFDM system with a DFT length of W and a CP length of N _C p . 1 The system is fully synchronized, and the channel's CIR (Channel Impulse Response) is h{P) (/ = 0, 1, N _ch -l , N _ch = maximum channel delay). After clearing the CP, the received time domain basis signal r(«) can be expressed as an OFDM symbol.

Where ") is a zero-mean Gaussian white-sounding sound (AWGN) and is independent of the transmitted signal · (").

By transforming w), r(n), and DFT, we get S(k), R(k), H(k), Z(k for this, we have

R(k) = H(k)S(k) + Z(k). 0≤k≤N-l (2) where point DFT is defined as

N-l .2mk

i

S(k) = DFT _N {s(n)} s(n)e N

Here we support a multipath propagation channel. The OFDM symbol duration is much larger than the channel's coherence time, which indicates that the CIR remains unchanged at least in adjacent symbols. This is an indisputable fact for LTE.

Now consider the received signal in the multiple sampling period 7; there is a time offset ^ and a normalized frequency offset CFO ε = NTsf _d (f _d = CFO [Hz]), then the received signal is

1=0

Note that here we omit the non-integer part of the time offset and the initial phase of the carrier, as they can be considered as part of the CIR. The coarse synchronization task is based on the received signal r(n) and some known pilot signals to estimate 0 and , ie, ^ and ^. Once the sum is calculated, BOS (or BOF) and CFO can be determined to determine the cell number (ID) and so on.

In LTE, there are two types of cyclic prefixes (CPs), namely regular CPs and extended CPs. In the case of a conventional CP, the subcarrier spacing is Δ/= 15 kHz, and when CP is extended, Δ/= 15, 7.5 kHz. Extended CPs are typically used in time-space environments to handle long channel delays (as in large cells). Af = 7.5 kHz for multimedia broadcasting (MBSFN) in single frequency networks. Although the time of one slot in LTE is fixed to 0.5 ms, the structure of the radio frame and the length of the CP are different. We assume that the DFT has a length of N when Δ/= 15 kHz.

FIG. 2 is a schematic diagram showing a frame structure of a 3GPP LTE FDD (Frequency Division Duplex) according to an embodiment of the present invention. Conventional CP, Af = 15 kHz: _ymb = 7 symbols per time slot (as shown in Figure 2). The first symbol has + ^^ sample points (the CP length is N _CPl ), and the other 6 symbols have N + Ne sample points (the CP length is N _CP .

Extended CP, Af = 15 kHz: = 6 symbols per slot, each symbol has N _symb = N + N _eCP samples (its CP length is N _eCP ,.

Extended CP, Af = 7.5 kHz: = 3 symbols per slot, each symbol has an equal symbol length N _symb = 2N + 2N _eCP (its CP length is 2N _eCP ).

In LTE, we have N _ECP > NCPI > N _CP2 . For a given sampling rate fs or sampling period T _s = l/f _s , the subcarrier spacing Δ /, the size of the DFT and the length of the CP can be confirmed Set. For example, a 5MHz bandwidth LTE, / _s = 7.68MHz, Δ = 15 kHz, we have W = 512, Nc i = 40, N _CP2 = 36, N _eCP = 128 „

In order to find the necessary information (such as the cell ID, PBCH in a radio frame, etc.) in the received signal, it is necessary to detect the length of the CP. Here, we propose a low complexity detection method based on lag self-correlation.

FIG. 3 shows a flow chart of a synchronous search method according to an embodiment of the present invention, including the following steps:

S102. Calculate a lag autocorrelation function of a data sample point of the received signal.

S104. Determine a time offset metric of the data sample point according to the lag autocorrelation function.

S106, calculating the time offset 0 of the data sample point and the estimated value of the frequency offset ε according to the time offset metric.

S108. Acquire a start point of the P-SCH symbol and a start point of the field according to the estimated value or self-correlate the received signal with the primary synchronization channel P-SCH.

In this embodiment, on the basis of the lag autocorrelation, the time-biased metric of the data sample point is determined, and then the time-biased metric is used to calculate the estimated value of the time-offset 0 and the frequency offset ε of the data sample point and according to the estimated value and The auto-correlation of the received signal with the primary synchronization channel P-SCH is performed to obtain the starting point of the P-SCH symbol and the starting point of the field, which reduces the complexity of the synchronous search, improves the efficiency, and overcomes the complexity of the synchronous search scheme in the prior art. Higher, more efficient than ^ 氐。.

Preferably, in the above method, calculating a lag autocorrelation function of the data sample point of the received signal comprises: determining an equivalent correlation window according to the available symbols of the received signal, and determining a lag autocorrelation function of the sampling point according to the equivalent correlation window;

.2πηε Ν^-\

Where r(n) = e ^N ^ h(I)s(n-0 + z(n), o≤n≤N-\ , _N

1=0

For the length of the discrete Fourier transform of the received signal, 0 and ε are the time offset and frequency offset of the data sample point r(n), respectively, 3⁄4(/) is the overall channel impulse response of the channel, / = 0, 1, N _ch - l, for the maximum channel delay, for the transmitted signal, ") is a zero-mean Gaussian Additive White Gaussian Noise (AWGN) independent of ^).

Preferably, in the above method, determining an equivalent correlation window according to available symbols of the received signal, and determining a lag autocorrelation function of the sampling point according to the equivalent correlation window includes:

C _AC (") = -^ ∑ ^r *( ^{w + m} ) ^r ( ⁿ + P + m) Where («) is a lagging autocorrelation function, which can be written in other forms, such as

1 ^w * w-\ *

C _AC (ή) = “∑ r(n + m)r (n + P + m) ^ C _AC (n) = ∑ r (n + m)r(n + P + m) , Temple.

W

W is the size of the sliding window, W=N _CP , W _CP is the length of the cyclic prefix of the symbols in the received signal, the corpse is hysteresis, P=N.

Preferably, in the above method, determining the time offset metric of the data sample point r(n) according to the lag autocorrelation function comprises: normalizing the lag autocorrelation function ^w)

C _AC (n)

P _AC ( ⁿ ) =

1 w i

Where ^W o , W _M=0 ; they can be written in other forms, mouth + P + z^l ² . Time bias metric

Λ(") = (")| , which can be written in other forms, such as A(n) = p _AC (n). When CP is used for detection

For BOS, it is usually possible to select corpse = W, = W _CP for DFT size, and ^ for CP length in one symbol). Metrics such as |C4 _C ( )| or |3⁄4 _c ( 7)| can be used to detect BOS and CFO. Preferably, in the above method, when there are multiple symbols available for receiving signals, the available symbols are determined according to the received signals, etc. The price-related window, and the lag autocorrelation function that determines the sampling points based on the equivalent correlation window includes:

_C {n+lN _symb +kN _sl

^syml slot 1=0 k=0 It can be written in other forms, ^; port CAc( ^{n + lN} symb ^{+ kN} slot)

Where n = 0, 1, N _symb -l, . , Is the number of time slots, _¾ of the number of symbols per _slot, N rfrt is the number of sampling points per slot, ^ is the number of sampling points per symbol, ^ C ( « )for,. ^ _rafr N The hysteresis correlation function of _CP sampling points.

Preferably, in the above method, determining the time offset metric of the data sample point r(n) according to the lag autocorrelation function comprises:

1

(^ = - "~― ∑ ∑E ₀ (n + lN _symb + kN _slot ) where K _symb K _slot / ₌₀ k=Q

1 -1 ― 1

E[{n) = - ~~—- ∑ ∑ E _x {n + IN _→ + kN _slot )

A symb A slot 1 = 0 k = 0 . Time skewness ΐ ^Λ (") = | ^(")|

Preferably, in the above method, calculating the time offset of the data sample point r(«) and the estimated value of the frequency offset ε according to the time offset metric ^ and includes:

Θ = arg max{A(«)} ε =—^C _AC (θ)

∑7t

The normalized correlation function (coefficient) does not depend on the energy of the signal and is therefore resistant to time-varying fading. Please note that the obtained data |3⁄4 _C («)| is sometimes called the relevant contour and is quasi-periodic. Within each symbol, a peak appears at the beginning of the CP. Especially when the current noise is low and there is no channel delay, max{A(«)}→l, the position of the peak is just the ideal BOS. For multipath channels, the peak will be delayed. The size of the delay is determined by the delay bandwidth of the channel. Since the size of the window W = N _CP is limited, the BOS thus obtained is very sensitive to channel and noise.

To this end, BOS can be sought by considering multiple symbols simultaneously. In general, when a symbol is available, we have a _CP equivalent to W _CP samples as the equivalent correlation window.

In simple terms, suppose there are _foi slots, each slot has ^ ₆ symbols. Each time slot has one sampling point, and each symbol has ^ sampling points, then the equivalent correlation window is fo Ncp sampling points, and the P-lag correlation can be written as

1 ^symb~^ K _s i _ot —\

In) =——- X 3⁄4 (n + lN _symb + kN _s (6)

^symb^slot 1=0 k=0 K

1 symb — ^{1 K} slot~^

∑ ∑^i (n + lN _symb

Nb丄, slot 1=0 k=0

Where w = 0, 1, N _symb —\. The metrics for BOS and CFO can be found in equation (9)

~(11) is used and replaced by C4 _C («) and 4 _C («). In calculating C4 _C («),

E ₀ (n), E^n) (n = 0 , 1, K _sl . _t K _symb N _symb — , we can calculate G "), Ε ₀ (η) , Ε (η) according to the above formula " = 0, 1, ..., N _symb —Y). It is worth noting that in this case all the peaks of C» are uniformly superimposed.

Preferably, in the above method, when the symbols are unequal length symbols, JV=N _{CP2 is set} , and any (W _CP i - W _CP2 ) connected sampling points are ignored or removed at the same position in each time slot.

For the case of unequal-length symbols (such as the regular CP in LTE), first, set the sliding window length W = N _CP2 , that is, the smaller CP length, and calculate C _AC according to the formula ( 5 ) ~ ( 8 ). ), E ₀ (n), E^n). The length of C _AC (n), E ₀ (n), thus obtained (ie, the range of values of its subscript n) is a time slot, each time slot There are Λ^. The sampling point is long. Any (W _CP i - NCP2) connected sample points are then ignored or removed at the same location in each time slot. Thus, the length of C _AC (n), E ₀ (n), O) becomes a time slot, and each slot N _slot — {N _{CP -} W _CP2 ) is sampled long. In other words, each time slot has ^ ^ = 7 symbols, and each symbol (W _CP2 + N) is long. By superimposing these values, you can get / («), then you can find BOS and CFO^.

4 shows a superimposed schematic diagram of metrics for a regular CP in LTE, in accordance with one embodiment of the present invention. Careful analysis will reveal that the BOS obtained in this case has some ambiguity even in a noise-free environment. In fact, for this, the actual BOS will be in the range of (Ne - W _CP2 ) (see Figure 4). If the ignored (W _CP1 - Ncra) samples are in the actual first symbol, then there is no ambiguity. When the ignored sample is not in the first symbol, the peak value of the BOS metric will deviate from its maximum deviation of {Na - NCP2) sample points. Because {Na - N _CP2 ) « N _CP2 , such a deviation is acceptable.

Preferably, in the above method, the method further includes the following steps:

Let W= N _CP2 , P = N, calculate the time-biased Λ0), where η = 0 , 1, ..., Detecting the width of the measure A(n), if the width has l~(N _ci - N _C P2 + 1) samples, the cyclic prefix is a regular cyclic prefix; otherwise, the cyclic prefix is an extended cyclic prefix; or

Width of the measure Λ(«) when detecting, if the width has more than one sample point, for example close to (Λ^ _Ρ -Λ^ _Ρ2 + 1) samples, the cyclic prefix is the extended cyclic prefix; otherwise, the cyclic prefix is regular Cyclic prefix.

Preferably, in the above method, the following steps may be further included: i W=N _CP2 , P = N, calculating a time-biased ΛΟ), where η = 0, 1, N _symh -\;

The peak metric «(«) of the detection time _CP ( _CP) , _max , if the peak value is greater than the set threshold, the cyclic prefix is the regular cyclic prefix; otherwise, the cyclic prefix is the extended cyclic prefix.

The size of the threshold is related to the definition of the time-biased metric, which can generally be determined by simulation or a small number of tests.

Preferably, in the above method, the method further includes the steps of: setting W=N _eCP , P = N, calculating a time-biased ΛΟ), where η = 0, 1, N _symh -\;

The peak value A _eCP , _max of the partial measure Λ(«) is _detected . If the peak value is greater than the set threshold, the cyclic prefix is the extended cyclic prefix; otherwise, the cyclic prefix is the regular cyclic prefix.

The size of the threshold can generally be determined by simulation or a small number of tests.

Preferably, in the above method, the following steps may be included:

Let W=N _CP2 , P = N, determine the peak value of the time-measure metric A _CP2 (w) A _CP2 , _max , let W=N _eCP , P = N, determine the peak value A _{eCP of the} time-measure metric A _eCP (w), _Max , where, " = 0, 1, ..., N _symb -l;

If A _C 2,max > A _eCP , leg _x , or A _C 2,max -. A _eC ,max> b («, b is a constant, which can be determined by simulation or trial), then the cyclic prefix is the regular cyclic prefix; otherwise, the cyclic prefix is the extended cyclic prefix.

Preferably, in the above method, the method further comprises the steps of: setting V=2N _eCP , P = 2N, calculating a time offset Λ(«), where η = 0, 1, ..., N _symb -l; The time offset Λ(«) peak A _2eCP , _max , if the peak A _2eCP , _{max is} greater than the set threshold, the length of the cyclic prefix

Otherwise, the length of the cyclic prefix is Δ/=15ΚΗζ. The size of the threshold can generally be determined by simulation or trial.

Detecting the CP type is equivalent to detecting different DFT sizes and CP lengths. These can be done with a support check. To check if the CP is Δ/=7.5ΚΗζ or

We can calculate the relevant contours within a symbol, such as = 2N _eCP for extended CPs, Corpse = 2W to calculate {A(w); w = 0, 1, N _symb —\,,. If ΛΟ) has a significant peak, then it is

Otherwise it is 15 baht. The reason is that for an inappropriate interval CP it is not possible to overlap with the sample it corresponds to within the symbol when making a correlation. Therefore, when there is a peak at the corpse = 2W, there is no peak at the corpse = W. On the contrary, when there is a peak at the corpse = W, there is no peak at the corpse = 2W.

To distinguish between regular CP and extended CP, we can set = N _C P2 , P = N and calculate the metric {A(w); w = 0, 1, ..., N _symb -Y}. Then we can detect the width of the metric peak. For conventional CP, the peak width is very narrow, with 1~(; W _CP1 -NCP2 + 1) samples. For example, at a sampling rate of 7.68 MHz, the peak width is 1 to 5 samples. But for extended CP, the peak width will be close (W _eCP - Ncp ₂ ) + 1 sample width. At a sampling rate of 7.68 MHz, this is equivalent to 85 samples. In addition, we can also use the parameters of the extended CP, W=N _eCP , P = N for detection. When a significant large peak is obtained, it is an extended CP, instead a regular CP (a regular CP has a small peak). The position of the peak is the required BOS.

It can be seen that by appropriately improving the lag autocorrelation method, we can simultaneously detect BOS, CFO and CP types. Knowing the CP type, the length of the OFDM symbol is also known accordingly.

In the published LTE Release 8 (release 8), 504 cellular numbers are defined, each number can be expressed as

=3 + ) ₍₉₎

Where N =0, 1, 167 is the ID of the cluster, N^=0, 1, 2 is the physical layer ID (sometimes called the sector ID) in a cluster.

The information of the physical layer IDN^ and the cluster ID N in the cluster is carried by two downlink SCH signals, namely, the primary synchronization channel P-SCH and the secondary synchronization channel S-SCH. Both P-SCH and S-SCH occupy 72 central subcarriers, and no signal propagates in direct current (DC) subcarriers in LTE (see Figure 2). Subframe 0 and subframe 5 of each radio frame contain P-SCH and S-SCH. The P-SCH is embedded in the last OFDM symbol, and the S-SCH is embedded in the second last symbol. For convenience, we divide the radio frame into two parts: the first half frame has sub-frames 0-4, and the second half frame consists of sub-frames 5-9. From the position of the P-SCH we can know the starting point of the field, but we cannot determine whether the starting point is the starting point of the radio frame (ie BOF). In LTE, N can be calculated in two steps:

(1) detecting which P-SCH is transmitted, you can get (2) The position of the S-SCH is determined based on the position of the P-SCH and N^, and then which S-SCH is transmitted is detected.

The initial synchronization signal P-SCH is selected from the odd-numbered Zadoff-Chu (ZC) sequence, and its signal in the frequency domain is defined as:

. uk ( ^ + 1 )

d _u {k) = e ^{3 Nzc} 0 ≤ k < N (10) where N _zc = 63 is the length of the ZC sequence, u is the ZC root index and has no common divisor greater than 1 with N _zc . A feature of the ZC sequence is that its autocorrelation is zero except for a maximum of one. This is a very good feature for synchronization. In the published LTE, the three physical IDs 0, 1, and 2 in the cluster are respectively represented by the P-SCH signal using three different ZC root indices u=25, 29, 34, so solving N^ means determining u.

In order to find the position of the P-SCH in the received signal, a cross-correlation operation is required between the received signal and the P-SCH signal. Cross-correlation operations are typically performed in the time domain. If the BOS is known, it can also be done in the frequency i or in the middle. Here we take the example of cross-correlation in time i or in the middle. First, three P-SCH time domain signals su(n) (u = 25, 29, 34) are generated. If the received signal is r(n) and the sliding window size is W = N, the cross-correlation coefficient can usually be defined as

C _xc (n, u)

(11)

Where 0^(",3⁄4) = "^2 :0) 0+∞) ( ₁₂ )

Its corresponding timing metric can be expressed as

Thus, the P-SCH position ή and the ZC root index ΰ can be obtained from

(η ) = argmax| 7 _c («,i)| ²

(16)

Similar to P-SCH, the secondary synchronization signal S-SCH also has only 62 non-zero sample values in the frequency domain. In short, the S-SCH signal is a binary sequence generated from N^ and N (see 3 GPP TS 36.211). In contrast to the P-SCH, the S-SCH signals are different from each other in subframes 0 and 5. In fact, this feature is used to block the molecular frames 0 and 5, thus identifying the starting point BOF of the radio frame.

Detecting the cell cluster ID N means detecting which S-SCH is embedded in the received signal. This can be determined by cross-correlating the S-SCH with the received signal.

In the absence of information about the OFDM symbol boundary, the coarse step timing is usually performed before the DFT, that is, in the time domain. Here, we briefly outline two timing methods for LTE: CP-based lag autocorrelation: As mentioned earlier, CP-based methods have low complexity, but can only detect CP types, BOS and CFO. The BOF and the cell IDs N^ and N^ require additional calculations:

Cross-correlation based on P-SCH: The cross-correlation based on P-SCH can determine the location of the P-SCH, so it can be used to calculate the starting point of the field and the BOS, and to determine the physical ID N^ in the cluster. In order to solve the problem of whether the P-SCH is in subframe 0 or 5, it is necessary to use the S-SCH. In addition, the frequency offset CFO cannot be obtained directly from the cross-correlation parameters and additional calculations are required.

The carrier frequency offset (CFO) can be divided into an integer frequency offset and a fractional frequency offset £ _F , ie £ = + ^(-1 < ^ < 1). All subcarriers are caused to shift in frequency Af, which results in interference between subcarriers. The fractional frequency offset can be calculated in the time domain by an autocorrelation based method. The integer frequency offset can be determined in the frequency i or. The details are as follows: First, determine the fractional frequency offset and corresponding compensation in the time domain, and then perform the FFT (assuming the symbol timing has been obtained before). Since the integer frequency offset is reflected by the cyclic shift of the signal in the frequency domain, it can be determined by the cross-correlation method and compensated accordingly.

Here, many synchronization methods can be combined to improve the computational uncertainty. For example, the detection of S-SCH can be used to check whether the obtained P-SCH is correct.

Cross-correlation-based methods usually have good performance, but they are computationally intensive, because metrics, such as | C _cx («, i ) | ² given in equation (12), need to be calculated for each sample _n . . On the other hand, it is a constant and can be calculated offline.

We notice energy, such as recursive calculations

WE _l (" + 1) = WE _l (") -

+ W)f (

For example, for an LTE system with a bandwidth of 20MHz (30.72 sample rate), we have W = N = 2408„ because El(n) and |r(n)|2 have been calculated before, and the equation (17) is calculated. El(n+1) in the middle only requires 3 MACs (multiply-accumulate multiplication port). This is equivalent to 3.30.72·106 - 93-106 MAC per second. This complexity of the algorithm is not difficult to achieve.

For cross-correlation, 1 calculates C _xc (", M) for each n and u requires 4-W MAC, ie 4.W.30.72.106 252.109 MAC per second. For 3 different u, total calculation The complexity will be as high as 756· 109 MAC per second. Such a computational complexity is generally not achievable with software solutions.

Conversely, the lag autocorrelation based on CP and P-SCH has low complexity, because ^(^ and ), ^w) can be recursively calculated.

WC _AC (" + 1) = WC _AC (n) - r* (n)r(n + P) + r* (n + W)r(n + P + W) (18) Note CAC(n) And r*(n)r(n+P) have been calculated before, therefore, only about 6 MACs are needed to calculate CAC(n+l). For the sampling rate of 30.72MHz, a total of 6.30.72·106 is required every second. 184-106MAC. Such a calculation is easy to solve today.

Preferably, in the above synchronous search method, the received signal is autocorrelated with the primary synchronization channel P-SCH according to the estimated value, and the starting point of the P-SCH symbol and the starting point of the field are obtained: the estimated value S^±(NCPl - At the NCP2) position, the received signal is cross-correlated with the P-SCH, and the position of the relevant contour peak is taken as the starting point of the P-SCH symbol; the starting point of the field is obtained by the starting point of the P-SCH symbol.

Preferably, in the above synchronous search method, the received signal is autocorrelated with the primary synchronization channel P-SCH according to the estimated value, and the starting point of the P-SCH symbol and the starting point of the field are obtained: the estimated value ^ ± (NCP1 - At the NCP2) position, the received signal is fast Fourier transformed and then cross-correlated with the P-SCH, and the position of the relevant contour peak is taken as the starting point of the P-SCH symbol; the starting point of the field is obtained by the starting point of the P-SCH symbol.

5 is a schematic diagram showing a P-SCH cross-correlation method according to the related art; FIG. 6 is a schematic diagram showing a start point of a synchronization method detecting a field according to an embodiment of the present invention; FIG. 7 is a diagram showing FIG. A comparison of performance comparison between the P-SCH cross-correlation of the embodiment and the synchronization method of the embodiment of the present invention.

In LTE, a 5 ms field has 10 subframes with a total of 70 OFDM symbols. Note that the starting point of the field is a BOS, but the BOS is not necessarily the starting point of the field. For all 70 BOS's neighborhoods, the position of the relevant contour peak is the starting point of the desired half-frame, which is also a more accurate BOS. It is noted that the P-SCH embedded in the received signal is also detected at this time, so the cell ID N is also known. In addition, this step is also available Executed in the frequency domain, that is, first FFT is performed on all BOS locations and their neighbors of the received signal, causing them to be converted to the frequency domain. The symbols in these frequency domains are then cross-correlated with the P-SCH in the frequency domain. The peak of its associated contour is at the beginning of the P-SCH symbol.

From the above description, it can be seen that the above-described embodiments of the present invention achieve the following technical effects:

Since the cross-correlation with P-SCH is not performed at all received signal sample points, but only at the location of the BOS and its neighbors (see Figure 6), the total computational complexity is greatly reduced. It is fully implemented with a software solution. According to our simulation, this low-complexity coarse synchronization scheme has almost the same superior performance as the P-SCH cross-correlation method (see Figure 7).

Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device, which can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Alternatively, they may be implemented by program code executable by the computing device, such that they may be stored in the storage device by the computing device, or they may be separately fabricated into individual integrated circuit modules, or they may be Multiple modules or steps are made into a single integrated circuit module. Thus, the invention is not limited to any particular combination of hardware and software.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the present invention. For those skilled in the art, the present invention may be variously modified and modified. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and scope of the present invention are intended to be included within the scope of the present invention.

Claims

Rights request

A synchronous search method, comprising the following steps: ^^

Calculating a lag autocorrelation function of the data sample points of the received signal;

Determining a time offset metric of the data sample point according to the lag autocorrelation function;

Calculating an estimate of the time offset ^ or the frequency offset ε of the data sample point according to the time offset metric or

£;

And obtaining a starting point of the P-SCH symbol and a starting point of the field according to the estimated value or autocorrelation of the received signal with the primary synchronization channel P-SCH.

2. The method according to claim 1, wherein calculating a hysteresis autocorrelation function of the data sample points of the received signal comprises:

Determining an equivalence correlation window according to the available symbols of the received signal, and determining a lagging autocorrelation function of the sample points according to the equivalent correlation window;

Wherein, the data sample point of the received signal is r{n) = (l)s(n - θ - 1) + z(n),

J

0 ≤ n ≤ N - \ , N is the length of the discrete Fourier transform of the received signal, and is the time offset and frequency offset of the data sample point r («), respectively, is the overall channel impulse response of the channel, 1 = 0, 1, ... , N _ch - 1 , N _ch is the maximum channel delay, which is the transmitted signal, which is a Gaussian whitening sound with an independent zero mean.

3. The method according to claim 2, wherein determining an equivalent correlation window according to an available symbol of the received signal, and determining a lag autocorrelation function of the sample point according to the equivalent correlation window comprises:

1 W-\

C _AC {ri) -― + m)r(n + P + m) , or

C _AC (n) = ∑ r (n + m)r(n --P --m) where C _AC (n) is the lagging autocorrelation function, which is the size of the sliding window, W=N _CP , N _C p is The length of the cyclic prefix of one symbol in the received signal, P is hysteresis, P=N.

4. The method according to claim 3, wherein determining the time offset metric of the data sample point r(«) according to the lag autocorrelation function comprises:

Normalizing the lag autocorrelation function C» C _AC (n)

E ₀ (n)E, (n) where,

Time-biased measure Λ(«) = | ^(«)

The method according to claim 3, wherein when the available symbols of the received signal are multiple, an equivalent correlation window is determined according to available symbols of the received signal, and according to the equivalent correlation window The post-correlation functions that determine the sample points are: lN _symb +kN _s

, or symt^slot 1=0 k=0

, ^symb~^ K _s i _ot —

C _A c(n)= ∑ ∑ C _AC (n + IN _b +kN _slot )

1=0 k=0

Where n = 0, 1, N _symb —\, ^ is the number of slots, ^^ is the number of symbols per slot, and N^ is the number of samples per slot, one For the number of samples of each symbol, C i) is the P-lag correlation function of K _slot K _symb N _CP m .

6. The method of claim 5, wherein determining the time offset metric of the data sample point r(«) according to the lag autocorrelation function comprises:

+lN _symb +kN _s J

or

^symlf slot 1=0 k=0

Ksyntb-l K _s i _ot —\

E ₀ {n)= ∑ ∑ E ₀ (n + 1N _b +kN _sIot )

1=0 k=0 E, (n + lN _{symb +} kN _slot )

, or

, ^symb~^ K _s i _ot —

E _l {n)= ∑ ∑ Ε _χ (n + lN _symb +kN _slot )

1=0 k=0 Time offset Λ( ) =

7. The method according to claim 4 or 6, wherein the time offset metric calculates an estimate of the time offset 0 and the frequency offset ε of the data sample point r(«) and comprises:

Θ = argmax{A(")}; ^_Z ^).

2π

The method according to any one of claims 1 to 7, characterized in that the unequal length symbol, or the type or length of the unequal length cyclic prefix, is determined by a method based on lagging autocorrelation.

9. The method according to claim 3, wherein when the symbol is a unequal length symbol, setting a correlation window to a shorter symbol cyclic prefix length = NGP2, the same in each of the time slots The position ignores or removes any (NC 1 - N _C P2 ) connected sample points, where NCP1 is the cyclic prefix length of the first symbol in the regular cyclic prefix.

10. The method according to claim 4 or 6 or 8, further comprising the steps of: W = N _CP2 , P = N, calculating a time-biased metric AO), wherein η = 0, 1 , N _symb -Y determines the width of the time offset Λ(«), if the width has 1 to (N _CT1 - N _C P2 + 1) samples, the cyclic prefix is a regular cyclic prefix, where ^^ is conventional The cyclic prefix length of the first symbol in the cyclic prefix;

Otherwise, the cyclic prefix is an extended cyclic prefix.

The method according to any one of claims 4 or 6 or 8, further comprising the steps of:

i .W = N _C p2 , P = N, the calculated time-biased ΛΟ), where η = 0 , 1 , N _{symb —} Y determines the peak A _CP2 , _{max of} the time-biased Λ(«) If the peak value is greater than a set threshold, the cyclic prefix is a regular cyclic prefix;

Otherwise, the cyclic prefix is an extended cyclic prefix.

Let W = N _eCP , P = N, calculate the time-biased measure Λ(«) , where η = 0 , 1 , N _symb —\ determine the peak value of the time-biased measure Λ(«) A _eCP , _max , if If the peak value is greater than a set threshold, the cyclic prefix is an extended cyclic prefix; Otherwise, the cyclic prefix is a regular cyclic prefix.

13. Method according to claim 4 or 6 or 8, characterized in that it further comprises the steps of: setting W = N _CP2 , P = N, determining the peak value A _CP2 , _{max of the} time offset metric A _CP2 («)

Provided _{W = N eCP, P = N} , the biasing metric A eCP ₍ «) peak A _eCT, _max is determined,

Where w = 0, 1 , N _symb - l ;

If Acp ₂ .max .max' or AC ₂ .max ― a A _eC pb {a, b is a constant), then the cyclic prefix is a regular cyclic prefix; otherwise, the cyclic prefix is an extended cyclic prefix.

The method according to any one of claims 8 or 10, further comprising the steps of:

Let W = 2N _eCP , P = 2N, calculate the deviation metric A(w), where w = 0, 1 , N _symb —\ determine the peak A _{2 of} the time-biased Λ(") _(? P,max If the peak A _2e c 'P, max is greater than a set threshold, the length of the cyclic prefix is Δ / = 7.5KHz;

Otherwise, the length of the cyclic prefix is Δ/ = 15ΚΗζ.

The method according to claim 14, wherein the received signal is autocorrelated with the primary synchronization channel P-SCH according to the estimated value, and the starting point of the P-SCH symbol and the starting point of the field are obtained:

And at a position of the estimated value ^ and ^ ± ( 0 _>1 - N _C p ₂ ), cross-correlating the received signal with the P-SCH, and using a position of a correlation contour peak as the P-SCH symbol The starting point of the field is obtained by the starting point of the P-SCH symbol.

At the estimated value and the position of ^ ±( 0 ₃₁ - N _cra ), the received signal is subjected to fast Fourier transform and cross-correlated with the P-SCH, and the position of the correlation contour peak is taken as the P-SCH The starting point of the symbol;

The start of the field is obtained by the start of the P-SCH symbol.