CN1665229A - Frequency Offset Estimation Method Based on Heterogeneous Spacing Frequency Domain Training Sequence - Google Patents

Abstract

The invention discloses a frequency bias estimating method based on a heterogenous-space frequency domain training sequence, a frequency synchronizing method for OFDM system and other blocked transmission systems, and the said training sequence is composed of inequi-distant frequency domain sequence and equidistant frequency domain sequence, where the energies of the two parts can be regulated but their total energy keeps invariant; besides, the said training sequence has a regulable length M; for resisting the interference between symbols, it inserts a circulating prefix with a length Ng in the front of the training sequence; the inequi-distant frequency domain sequence is composed of MD inequi-distant pilot frequencies; the equidistant frequency domain sequence is composed of MU equidistant pilot frequencies and generated by a ''Chu'' sequence with circulating orthogonal property; the method is fast and reliable, has relative small load and large capturing range, ability to resist multipath effect, high estimating accuracy, and low implementing complexity, both suitable for continuous data transmission and data packet transmission.

Description

Frequency Offset Estimation Method Based on Heterogeneous Spacing Frequency Domain Training Sequence

                      

The invention relates to a frequency synchronization method applied to an OFDM (orthogonal frequency division multiplexing) system and other block transmission systems, and belongs to the technical field of synchronization in mobile communication.

Background technique

Frequency synchronization is a prerequisite for normal communication in a mobile communication system. In order to support high-speed data services, the future mobile communication system will be a broadband, multi-antenna system, and OFDM is an important candidate for the future mobile communication system. For future mobile wireless communications, the time-varying characteristics of broadband wireless channels will affect the carrier frequency and cause it to shift, thus destroying the orthogonality between subcarriers in the OFDM system. Compared with single-carrier systems, OFDM systems are more sensitive to carrier frequency offset. How to reduce the impact of ICI on system performance is one of the prerequisites for OFDM systems to be widely used. Traditional frequency synchronization methods are based on frequency domain training sequence or time domain training sequence to estimate the carrier frequency offset, they all have such and other shortcomings: not suitable for packet data transmission, high load, small capture range, It cannot resist multipath, the estimation performance is not ideal, and the calculation complexity is high.

Contents of invention

Technical problem: The purpose of the present invention is to overcome the above shortcomings, provide a frequency-domain training sequence structure with adjustable parameters, and further provide a fast and reliable, relatively small load, large capture range, and anti-multipath effect. A frequency offset estimation method based on heterogeneous spacing frequency-domain training sequences, which has high estimation accuracy, low implementation complexity, and is suitable for both continuous data transmission and packet data transmission.

Technical solution: The heterogeneously spaced frequency domain training sequence structure with adjustable parameters of the present invention is mainly composed of an unequally spaced frequency domain sequence comprising M _D unequally spaced pilots and an equally spaced frequency domain sequence comprising M _U equispaced pilots Sequence composition; in order to combat ISI (Inter-Symbol Interference), a cyclic prefix of length N _g is inserted in front of this training sequence. Under the constraint that the energy ξ of the entire training sequence is constant (ξ=M), the ratio of the energy ξ ₁ and ξ ₂ of the unequal-spaced frequency-domain sequence and the equally-spaced frequency-domain sequence, that is, ξ ₁ : _{ξ 2} = α, can be adjustment; at the same time, the length M of the entire sequence, and the numbers M _D and M _U of the frequency domain sequences included in the two parts of the sequence are also adjustable. By selecting appropriate parameters, the corresponding frequency offset estimators can obtain different performances, so that they can be applied to different wireless mobile scenarios.

The carrier frequency offset estimation method based on the parameter-adjustable heterogeneous spacing frequency domain training sequence of the present invention is:

(1) Estimate the rough frequency offset according to the received time-domain sequence, and obtain the corresponding frequency-domain sequence;

(2) Set the unequal spacing sequence and adjacent subcarriers contained in the above frequency domain sequence to zero;

(3) Transform the frequency domain sequence obtained in (2) into the time domain for coarse frequency offset correction;

(4) Perform fine frequency offset estimation based on the above-mentioned time domain sequence after coarse frequency offset correction;

(5) Add the estimated coarse frequency offset and fine frequency offset to get the total estimated frequency offset.

Among them, the method of rough frequency offset estimation based on the received time domain sequence, its specific implementation steps are:

(1) Calculate its periodogram according to the received time domain sequence;

(2) Use the bubble method to search for the peak amplitude of the corresponding periodogram;

(3) Determine the index value of the found peak pilot in the set {d _k } ₀ ^MD-1 according to the lookup table;

(4) Calculate the offset of the found peak frequency domain pilot and normalize it to N, so as to determine the rough frequency offset estimate.

Among them, the set {d _k } ₀ ^MD-1 represents the index values of M _D unequal-spaced pilots in the heterogeneous-spaced frequency-domain training sequence; N is the number of subcarriers contained in one OFDM symbol.

Described look-up table, its storage content is as follows: d ₁ -d ₀ d ₂ -d ₀ … d _MD-1 -d ₀ d ₂ -d ₁ d ₃ -d ₁ … d ₀ -d ₁ +M     d ₀ -d _MD-1 +M d ₁ -d _MD-1 +M … d _MD-2 -d _MD-1 +M

Wherein, d ₀ , d ₁ , ..., d _MD-1 are index values of M _D unequal-spaced pilots in the heterogeneous-spaced frequency-domain training sequence.

Beneficial effect:

1. Introducing the concept of heterogeneously spaced frequency domain training sequences, making full use of the respective advantages of unequally spaced frequency domain sequences and equally spaced frequency domain sequences, so as to obtain the optimal estimation performance.

2. Using a lookup table, making full use of the structural characteristics of frequency-domain training sequences with unequal energy heterogeneous spacing, improves the correct probability of coarse frequency offset estimation, and greatly reduces the time consumption for completing coarse frequency offset estimation; The equally spaced frequency domain sequence obtained by intersecting the time domain sequence also greatly improves the accuracy of the fine frequency offset estimation algorithm.

3. According to the actual carrier frequency offset and specific application scenarios, frequency domain training sequences with different lengths are selected to obtain the optimal load performance trade-off.

The frequency offset estimation algorithm proposed by the invention can be used in any block transmission system.

The invention mainly considers how to reduce the system load in the mobile communication system, reduce the complexity of the estimation algorithm, improve the system performance, and enable the system to support high-speed data services more efficiently.

Description of drawings

FIG. 1 is a schematic diagram of the structure of the heterogeneous spacing frequency domain training sequence of the present invention. Among them, d ₀ , d ₁ , d ₂ , d ₃ ,..., d _MD-2 , d _MD-1 are the index values of M _D unequal-spaced pilots in the heterogeneous-spaced frequency-domain training sequence; u ₀ , u ₁ , u ₂ , u ₃ ,..., u _MU-2 , u _MU-1 are the index values of M _U equally spaced pilots in the heterogeneously spaced frequency domain training sequence; c ₀ , c ₁ , c ₂ , c ₃ , ... , c _MC-2 , c _MC-1 are the index values of M _C non-zero pilots in the heterogeneous spacing frequency domain training sequence; M _C = M _D + M _U ; M is the total length of the heterogeneous spacing frequency domain training sequence; ξ ₁ , ξ ₂ , ξ are the energy values of unequal-spaced pilots, equal-spaced pilots and non-zero pilots in heterogeneously spaced frequency-domain training sequences, respectively.

Fig. 2 is a schematic diagram of a frequency offset estimation method based on heterogeneous spacing frequency domain training sequences.

FIG. 3 is a schematic diagram of an implementation structure of a frequency offset estimation algorithm based on heterogeneous spacing frequency domain training sequences. It includes a multiplexing device, an FFT (fast Fourier transform) device, a square device, a peak amplitude search device, a peak pilot index calculation device, an offset calculation and a normalization device, a conjugate device, a zeroing device, a multiplication device, Multiplication and accumulation device, phase angle calculation device, and addition device.

Detailed ways

Assume that the number of subcarriers contained in an OFDM symbol is N, the length of the cyclic prefix is N _g , and the length of the heterogeneously spaced frequency-domain training sequence is M, which consists of M _D unequal-spaced pilots and M _U equal-spaced pilots. composition. Among them, the index of the unequal-spaced frequency domain sequence is represented by {d _k } ₀ ^MD-1 , and the index of the equidistant frequency-domain sequence is represented by {u _k } ₀ ^MU-1 ; the ratio of the overall energy of the two pilots is set to ξ ₁ : ξ ₂ = α. Let M _C = M _D + M _U , then the above non-zero pilot sequence can be expressed as ${\tilde{p}}_{m_C}={[{\tilde{p}}_{c_0},{\tilde{p}}_{c_1},\&Center\; Dot;\&Center\; Dot;\&Center\; Dot;{\tilde{p}}_{c_{m_C-1}}]}^T.$

After the received time domain sequence undergoes FFT operation, calculate its periodogram and perform peak amplitude search, and then pass through the peak pilot index value calculation unit, then calculate the offset of the peak pilot and normalize it to N to obtain Corresponding coarse frequency offset estimation value, and then according to the obtained rough frequency offset estimation value, the unequally spaced pilots and adjacent subcarriers contained in the above frequency domain sequence are set to zero, and then the IFFT (inverse Fourier transform) operation is performed to zero After the sequence is converted to the time domain, and coarse frequency offset correction is performed; the corrected time domain sequence is sent to the fine frequency offset estimation unit to obtain the fine frequency offset estimation value; finally, the coarse and fine frequency offset estimation values are added together , output the total frequency offset estimate.

The specific algorithm is described as follows:

The expression of the received time domain sequence affected by the normalized frequency offset ε can be written as:

[formula one]

where φ is the phase deviation introduced due to timing error or Wiener phase noise, $\stackrel{\‾}{f}=[f_{c_0},f_{c_1},\·\·\&Center\; Dot;,f_{c_{m_C-1}}]$ is the IFFT transformation matrix of M×M _C , and w is the additive Gaussian white noise signal. Then, the periodogram of the received sequence is calculated by FFT operation:

$\mathrm{\ξ}\left(k\right)={\left|\underset{\mathrm{no}=0}{\overset{m-1}{\mathrm{\Σ}}}{r}_{\mathrm{no}}{e}^{-j2\mathrm{\πnk}/m}\right|}^2,k=\mathrm{0,1},\·\·\·,m-1$ [formula two]

Send the above signal to the peak amplitude search unit to find the maximum value as follows:

$\mathrm{\ζ}=\underset{k\∈[0,m-1]}{\arg \max }\left\{\mathrm{\ξ}\left(k\right)\right\}$ [formula three]

Then, locate the index value of the peak pilot signal in the set {d _k } ₀ ^MD-1 according to the lookup table, namely:

$\mathrm{\κ}=\underset{k[0,m_{\mathrm{D.}}-1]}{\arg \max }\left\{\underset{g=0}{\overset{m_{\mathrm{D.}}-2}{\mathrm{\Σ}}}\mathrm{\ξ}\right[{\left(({\mathrm{\Π}}_{k,g}+\mathrm{\ζ})\right)}_m\left]\right\}$ [formula four]

Wherein, II _{k, g} represent the content stored in row k and column g in the lookup table. Send the result of the above formula to the offset calculation and normalization module to get the rough frequency offset estimate.

$\mathrm{\δ}=\frac{N}{m}(\mathrm{\ζ}-d_k)$ [formula five]

[formula six]

Send the frequency domain sequence obtained after the FFT operation to the zeroing module to perform the zeroing operation according to the following formula,

[公式七]

[formula seven]

Then, the zero-set frequency domain sequence is converted to the time domain through the IFFT operation, and the time domain sequence is sent to the corresponding coarse frequency offset correction module:

$r_{\mathrm{no}}^{\mathrm{cc}}=r_{\mathrm{no}}^{\mathrm{ic}}{e}^{-j2\mathrm{\π}{\widehat{\mathrm{\ϵ}}}_c\mathrm{no}/N},\mathrm{no}=\mathrm{0,1},\·\·\·,m-1$ [formula eight]

Send the corrected time domain sequence to the fine frequency offset estimation module in turn to obtain the fine frequency offset estimation value as follows:

[formula nine]

Among them, X=M/M _U ;

Finally, send the estimated coarse and fine frequency offset estimates to the adder to obtain the total frequency offset estimate as follows:

[formula ten]

According to the above description, the implementation steps of the frequency offset estimation algorithm based on the heterogeneous spacing frequency domain training sequence can be obtained as follows:

(1) Perform rough frequency offset estimation according to the received time-domain sequence, and obtain the corresponding frequency-domain sequence;

(2) Set the unequal spacing sequences and adjacent subcarriers contained in the frequency domain sequence to zero;

(3) Transform the frequency domain sequence obtained in (2) into the time domain for coarse frequency offset correction;

(4) Perform fine frequency offset estimation according to the time domain sequence after the coarse frequency offset correction;

(5) Add the estimated coarse frequency offset value and the fine frequency offset value to obtain the total estimated frequency offset value.

Among them, the coarse frequency offset estimation algorithm based on heterogeneous spacing frequency domain training sequence, its specific implementation steps are:

(1) Calculate its periodogram according to the received time domain sequence;

(2) Use the bubble method to search for the peak amplitude of the corresponding periodogram;

(3) Determine the index value of the found peak pilot in the set {d _k } ₀ ^MD-1 according to the lookup table;

(4) Calculate the offset of the found peak pilot and normalize it to N, so as to determine the rough frequency offset estimate.

Among them, the set {d _k } ₀ ^MD-1 represents the index values of M _D unequal-spaced pilots in the heterogeneous-spaced frequency-domain training sequence; N is the number of subcarriers contained in one OFDM symbol.

The FFT device and the square device complete the calculation contained in the formula [2], the peak amplitude search device completes the formula [3], the peak pilot index value calculation device completes the formula [4], and the offset calculation and normalization device completes the formula [4]. Five], formula [six], the zero setting device completes the formula [seven], the conjugate device and the FFT device complete the IFFT operation, the multiplication device completes the coarse frequency offset correction operation (formula [eight]), two multiplication accumulation devices, and the multiplication device , and the phase angle calculation device are combined to complete the fine frequency offset correction operation (formula [9]), and the addition device completes the total frequency offset estimation operation (formula [10]).

Claims (2)

1. A frequency offset estimation method based on heterogeneous spacing frequency domain training sequences, characterized in that: the energy of the two parts of the training sequence composed of unequal spacing frequency domain sequences and equal spacing frequency domain sequences can be adjusted, but The total energy remains unchanged; at the same time, the length M of the heterogeneously spaced frequency-domain training sequence is also adjustable; in order to combat inter-symbol interference, a cyclic prefix of length _Ng is inserted in front of the training sequence; the unequal-spaced frequency-domain sequence It consists of M _D pilots with unequal spacing; the equally spaced frequency domain sequence is composed of M _U pilots with the same spacing, and is generated by the "Chu" sequence with cyclic orthogonality; The frequency offset estimation method is: Step 1), perform rough frequency offset estimation according to the received time-domain sequence, and obtain the corresponding frequency-domain sequence; Step 2), setting the unequal spacing sequences and adjacent subcarriers included in the frequency domain sequence to zero; Step 3), transforming the frequency domain sequence obtained in step 2) into the time domain for coarse frequency offset correction; Step 4), performing fine frequency offset estimation according to the above-mentioned time domain sequence after coarse frequency offset correction; Step 5), adding the estimated coarse frequency offset value and fine frequency offset value to obtain a total estimated frequency offset value. 2. The frequency offset estimation method based on heterogeneous spacing frequency domain training sequence according to claim 1, characterized in that the method for rough frequency offset estimation according to the received time domain sequence is: according to the periodogram of the received time domain sequence, Frequency offset estimation is performed based on a lookup table, and the specific implementation steps are as follows: (1) Calculate its periodogram according to the received time domain sequence; (2) Use the bubble method to search for the peak amplitude of the corresponding periodogram; (3) Determine the index value of the found peak pilot in the set {d _k } ₀ ^MD-1 according to the lookup table; (4) Calculate the offset of the found peak pilot and normalize it to N, thereby determining the rough frequency offset estimate; Among them, the set {d _k } ₀ ^MD-1 represents the index values of M _D unequal-spaced pilots in the heterogeneous-spaced frequency-domain training sequence; N is the number of subcarriers contained in one OFDM symbol.

Images