CN1398016A - Method for generating 3D wave beams in intelligent antenna - Google Patents

Abstract

This invention provides a method of forming three dimentional beams with an array antenna composed of x-y plane uniform round array and Z shaft linear array (or a linear array of x-y-z three linears vertical to each other) to divide the three dimentional sphere into uniform gratings and find out related lattice points according to the DOA of the given dexpected signals to be given the weight number of per-unit value, while the weight numbers at the central places of other gratings are zero to form a group of restrict conditions then to realize a method of spreading MMSE beams with restrict conditions with the MMSE method.

Description

A kind of three-dimensional Beamforming Method in the smart antenna

Technical field

The invention belongs to communication technique field, be specifically related to the formation method of three-dimensional wave beam in a kind of smart antenna.

Background technology

In recent years, using smart antenna raising power system capacity and reduce disturbance in the radio communication has been an important topic. Adaptive beam in the receiver of smart antenna forms the Signal to Interference plus Noise Ratio that can improve in the Radio Link, reduce channel block, increase total power system capacity ([1] Jr Liberti J C, Rappaport T S.Smart Antennas for Wireless Communications:IS-95 and Third Generation CDMA Applications. Upper Saddle River, NJ, Prentice Hall PTR, 1999, [2] Razavilar J, Rashid-Farrokhi F, Liu Ray K J.Software radio architecture with smart antennas:A tutorial on algorithms and complexity.IEEE Journal on Selected Areas in Communications, 1999,17 (4), pp.662-675.), forming wave beam by weighting makes the approximate zero gain direction aim at multipath signal and high reject signal source, strong gain direction is aimed at desired signal source, can reduce to disturb, reduce user emission power, the shared same channel of a plurality of mobile subscribers and base station are communicated. Because these array weights form the operation of wave beam and can finish in intermediate frequency or base band, therefore can realize with DSP.

Some documents in the past all are the formation of research 2 dimensional plane wave beams, this obviously can not with reality in 3 dimension spaces adapt, so formation ([3] Nehorai A of some literature research 3 dimension wave beams just occurred, Ho K C, Tan B T G.Minimum-Noise-Variance Beamformer with an Electromagnetic Vector Sensor.IEEE Transactions on Signal Processing, March 1999,47 (3), pp.601-618.), because 3 dimension spaces are more more wide than 2 dimensions, therefore more original Beamforming Methods such as MVDR, MMSE, LS, MSNR, adaptive algorithm, the formed wave beam of blind adaptive algorithm are always with very large secondary lobe, and main lobe is very wide, the method that wherein has such as the weight coefficient of MMSE method on the desired signal direction do not reach maximum, the method that has such as LS method then can not effectively suppress to disturb and noise, the output Signal to Interference plus Noise Ratio is low, and namely said method all can not reach best.

Summary of the invention

The object of the invention is to propose a kind of can establishment the interference and noise, have high output Signal to Interference plus Noise Ratio, can make again the wave beam main lobe of formation narrower, the formation method of 3 dimension wave beams in the less smart antenna of secondary lobe.

Three-dimensional Beamforming Method in the smart antenna that the present invention proposes, to adopt by x-y uniform plane circle battle array and z axle line array or the array antenna received signals (hereinafter getting for the time being front a kind of array antenna) that formed by three orthogonal line arraies of x-y-z, and with the even lattice of 3 n-dimensional sphere ns, DOA according to known desired signal finds out corresponding lattice point, and tax is with the weight coefficient of unit value, it is 0 that weight coefficient on the center of all the other grid is then composed, form restriction condition, unite to obtain with the MMSE method and form the required weight vector of wave beam, so claim that the present invention is the expansion MMSE Beamforming Method with restriction condition.

Simulation shows, the formed wave beam of weighing vector of being tried to achieve by the inventive method, as shown in Figure 2, not only main lobe is narrower, and secondary lobe is less, and the weight coefficient on the desired signal direction is the highest, reach 1, can effectively suppress again to disturb and noise, so that the output Signal to Interference plus Noise Ratio of Beam-former is higher, for example comparable least square (LS) Beamforming Method increases more than 50 times.

The below is described further the inventive method.

With as shown in Figure 1 the array antenna that is comprised of x-y uniform plane circle battle array and z axle line array, the array element of line array is spaced apart d when forming wave beam₁, circle battle array radius is d₂, always (M can be selected by oneself for total M array element. In principle, more the wave beam that becomes of multiform is better for M, but system architecture is also complicated, considers both balances, general M desirable 10～20), each is omnidirectional's oscillator, the guiding vector of this array antenna is $a(\mathrm{\φ},\mathrm{\ψ})=[e^{{\mathrm{j\ω\τ}}^1}{e}^{\mathrm{j\ω}{\mathrm{\τ}}^2}...e^{{\mathrm{j\ω\τ}}^M}{]}^T---\left(1\right)$ φ wherein, ψ is azimuth and the elevation angle of arrival bearing DOA, ω is carrier angular frequencies, τ^m＝u(φ，ψ)r _m/c， m∈{1，2，…，M}，u(φ，ψ)＝[cosφcosψ，sinφcosψ，sinψ]，r _m＝[x _m y _m z _m] ^TBe the space coordinates of m array element, c is the light velocity.

Suppose to receive K CDMA user's signal, each user has L_k(k=1,2 ..., K) bar multipath. Channel parameter can be regarded LTI (LTI) as in short time slot, make α_k, p is the attenuation coefficient of k user's p bar multipath, below for sake of convenience, will send amplitude

With carrier phase _kBe also contained in α_k，pIn, when information source and receiver carrier synchronization_k=0, so α_k，pSignal amplitude for the array element reception. τ_k，pIt is the arrival time delay (TOA) of k user's p bar multipath. Getting the sampling period is the spread-spectrum code chip cycle T_c, establish sampling time T_sMuch smaller than T_c, therefore can think point sampling. The array received signal vector of n sampling instant then $r\left(n\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{p=1}{\overset{L_k}{\mathrm{\Σ}}}a({\mathrm{\φ}}_{k,p},{\mathrm{\ψ}}_{k,p})a_{k,p}{c}_k^{\left(\right[\left({n-c}_{k,p}\right)/N\left]\right)}{a}_k^{\left(\mathrm{mod}({n-v}_{k,p},N)\right)}+n\left(n\right),---\left(2\right)$ Wherein T is symbol intervals, N=T/T_cBe spreading gain, $\left\{c_k^{\left(i\right)}\right\},c_k^{\left(i\right)}\∈\{1,-1\}$ Be the symbol that k user sends,Be k user's spread spectrum code sequence, for BPSK (binary phase shift keying) modulation, $a_k^{\left(n\right)}\∈\{1,-1\}.$ White Gaussian noise vector when n (n) is the n time sampling, the average of its each yuan is 0, variance is

SymbolExpression rounds, $v_{k,p}=\left[\mathrm{mod}({\mathrm{\τ}}_{k,p}/T_c,N)\right]$ Be integral multiple T_cThe arrival time delay. φ_k，p、ψ _k，pBe azimuth and the elevation angle of the 7th user's p bar multipath. Get a segment length and be N sequence observing time [iN, iN+1 ..., iN+N-1], corresponding to a symbol intervals T. Output signal vector is during this period of time formed an observation matrix

          Y＝[r(iN)，r(iN+1)，…，r(iN+N-1)] ^T。        (3)

Be without loss of generality, if need to form the wave beam of the 1st paths of aiming at the 1st user, then see other each road signals as interference. Sphere evenly is divided into N₁×N ₂(General Requirements N₁×N ₂＞M，N ₁Desirable 10～20, N₂Desirable 5～10) individual grid, angle group (azimuth, the elevation angle) corresponding to its center is as follows:Form (a N by the Θ matrix₁×N ₂The matrix of) * M $A=[a\left({\mathrm{\θ}}_{\mathrm{1,1}}\right)a\left({\mathrm{\θ}}_{\mathrm{1,2}}\right)...a\left({\mathrm{\θ}}_{1,N_2}\right)a\left({\mathrm{\θ}}_{\mathrm{2,1}}\right)...a\left({\mathrm{\θ}}_{N_1,N_2}\right)]---\left(4\right)$ θ wherein_i，jThe element that lists for the capable j of i of matrix Θ. Again definitionBe expectation response vector, the N of corresponding Θ₁×N ₂Group angle θ_1，1 θ _1，2 … θ _2，1… , in Θ, find out the DOA with the 1st user's who estimates the 1st paths(its method of estimation can have multiple, better can adopt 3 array element DOA methods of estimation---see another patent of inventor) immediate angle group, then be 1 with element definition corresponding with this angle group among the z, all the other elements all are defined as 0, obtain restricting equation: w^HA=z ^H。

For suppressed sidelobes to greatest extent, when guaranteeing that main lobe is narrower, can obtain again the maximum weighted coefficient on the desired signal direction, with A, z expands to respectively A_e，z _e：A _e=[A a(φ _1，1，ψ _1，1)]， $z_e=\left[\begin{array}{c}z\\ 1\end{array}\right],$ Form new restriction equation: w^HA _e＝z _e If adding the MMSE method of this restriction condition namely expands output signal that the MMSE method makes Beam-former and approaches training sequence d (by the derivation of back as can be known, do not need this training sequence when calculating weight vector at last), so the weight vector of expansion MMSE method is $w=\arg \underset{w}{\min }E\left[\right|d\left(n\right)-w^{H}r\left(n\right){|}^2]s.t.w^H{A}_e=z_e^T---\left(5\right)$ Find the solution with the Lagrange multiplier method, make object function be $L\left(w\right)=E\left[\right|d\left(n\right)-w^{H}r\left(n\right){|}^2]+\underset{i=1}{\overset{N_1\×N_2+1}{\mathrm{\Σ}}}{\mathrm{\λ}}_i{w}^H{A}_e(:,i)$ A _e(:, i) A is got in expression_eI row, object function is asked local derviation: $\frac{\∂L\left(w\right)}{\∂w}=2E\left[r\left(n\right)r^H\left(n\right)\right]w-2E\left[r\left(n\right)d^{*}\left(n\right)\right]+2A_e\mathrm{\Λ}$ Wherein $\mathrm{\Λ}=[{\mathrm{\λ}}_1,{\mathrm{\λ}}_2,...,{\mathrm{\λ}}_{N_1\×N_2+1}{]}^T$ Make that following formula is 0, the weight vector of the MMSE method that is expanded $\hat{w}=R^{-1}(p-A_e\mathrm{\Λ}),$ R=E[r (n) r (n) wherein^H]，p＝E[r(n)d ^*(n)]. Substitution restriction equation: $A_e^H{R}^{-1}(p-A_e\mathrm{\Λ})=z_e$ So $p-A_e\mathrm{\Λ}={\left(A_e^H{R}^{-1}\right)}^{-1}{z}_e.$ For larger spreading gain, R=Y can be arranged^HY, so $\hat{w}={\left(Y^{H}Y\right)}^{-1}{\left(A_e^H{\left(Y^{H}Y\right)}^{-1}\right)}^{-1}{z}_e---\left(6\right)$ It is right to need before asking this weight vector $A_e^H{\left(Y^{H}Y\right)}^{-1}$ Carry out singular value decomposition (SVD), obtain its U, V and ∑ matrix, make $A_e^H{\left(Y^{H}Y\right)}^{-1}=\mathrm{U\Σ}{V}^{*},$ Wherein ∑ is diagonal matrix, and U, V are respectively $A_e^H{\left(Y^{H}Y\right)}^{-1}$ Left singular matrix and right singular matrix. The weight vector of MMSE method is so be expanded at last $\hat{w}={\left(Y^{H}Y\right)}^{-1}{\mathrm{V\Σ}}^{-1}{U}^{-1}{z}_e---\left(7\right)$

Description of drawings

Fig. 1 is the array antenna diagram that is comprised of x-y uniform plane circle battle array and z axle line array.

Fig. 2 is the 3 dimension wave beam diagrams that form with the inventive method.

The specific embodiment

Further specifically describe the present invention below by examples of implementation. Suppose to receive altogether K=2 user's signal, each user has L_kArticle=5, multipath. Each user's spreading code length is 255, and correlation is relatively good. It is 1 training sequence that the user sends symbol, noise variance ${\mathrm{\σ}}_n^2=0.01.$

1, choose at random the antenna reception amplitude of 10 road signals during simulation: α=rand (10) sees Table 1.

2, during simulation [0, choose at random the arrival time delay of 10 road signals: τ=NT between T)_c×rand(10)。

3, during simulation (π, π] between choose at random the azimuth of 10 road signals: φ=2 π * rand (10)-π; $[-\frac{\mathrm{\π}}{2},\frac{\mathrm{\π}}{2}]$ Between choose at random the elevation angle of 10 road signals: ψ=π * rand (10)-pi/2.

The array antenna of 4, Fig. 1 always has 13 array elements, and wherein z axle line array has 5 array elements, and x-y uniform plane circle battle array has 8 array elements, and carrier wavelength is got at the array element interval of circle battle array radius and line array half:

d ₁＝d ₂＝λ/2。

5, N when sphere is divided grid₁、N ₂Obtain greatlyr, the wave beam of formation must be better to Sidelobe Suppression, and main lobe width is narrower, but operand is also larger, gets N here₁＝20，N ₂=10. Analog result sees Table 1 and Fig. 2.

Table 1: the output amplitude of Beam-former after the actual reception amplitude of the antenna of getting at random and the weighting

User k	Path p	α k，p	MMSE		LS		Expansion MMSE
									β k，p	G k，p	-β k，p	G k，p	β k，p	G k，p
			1     2	1     2     3     4     5     1     2     3     4     5	0.8971   0.7965   0.7550   0.6133   0.5081   0.8428   0.8205   0.7107   0.7055   0.6023	0.8687   0.0009   0.0085   0.0695   0.0409   0.0452   0.1789   0.073l   0.0096   0.0272	0.9683   0.001l   0.0112   0.1134   0.0805   0.0536   0.2180   0.1029   0.0137   0.045l	0.8847   0.0004   0.0063   0.0942   0.0684   0.0466   0.2289   0.1068   0.0184   0.0267							0.9861   0.0005   0.0083   O.1536   0.1347   0.0553   0.2790   0.1503   0.0261   0.0443	0.8971   0.0021   0.0033   0.0016   0.0004   0.0032   0.0020   0.0018   0.0014   0.0013	1.0000   0.0027   0.0043   0.0026   0.0008   0.0038   0.0024   0.0025   0.0020   0.0022
SINR		(0.175l input)							(17.182 output)		(9.5694 output)		(625.00 output)

β _k，p＝α _k，pw ^Ha(φ _k，p，ψ _k，p) be k user's p bar multipath signal through the weighted amplitude behind the Beam-former: $G_{k,p}=\frac{{\mathrm{\β}}_{k,p}}{{\mathrm{\α}}_{k,p}}=w^{H}a({\mathrm{\φ}}_{k,p},{\mathrm{\ψ}}_{k,p})$ Be the weight coefficient of Beam-former to k user's p bar multipath signal.

Claims (3)

1, the formation method of three-dimensional wave beam in a kind of smart antenna, it is characterized in that adopting by x-y uniform plane circle battle array and z axle line array or the array antenna received signals that formed by three orthogonal line arraies of x-y-z, and with the even lattice of 3 n-dimensional sphere ns, DOA according to known desired signal finds out corresponding lattice point, and tax is with the weight coefficient of unit value, it is 0 that weight coefficient on the center of all the other grid is then composed, form restriction condition, unite with the MMSE method and obtain the required weight vector of formation wave beam.

2, according to three-dimensional Beamforming Method in the described smart antenna of claim l, it is characterized in that the weight vector that forms wave beam is obtained by following formula: $\hat{w}={\left(Y^{H}Y\right)}^{-1}V{\mathrm{\Σ}}^{-1}{U}^{-1}{z}_e---\left(7\right)$ Y=[r (iN) wherein, r (iN+1) ..., r (iN+N-1)^T；                         (3)

U, V and ∑ matrix satisfy $A_e^H{\left(Y^{H}Y\right)}^{-1}=\mathrm{U\Σ}{V}^{-1}$ , wherein ∑ is diagonal matrix, U, V are respectively $A_e^H{\left(Y^{H}Y\right)}^{-1}$ Left singular matrix and right singular matrix; $A_e=\left[\mathrm{Aa}({\mathrm{\φ}}_{\mathrm{1,1}},{\mathrm{\ψ}}_{\mathrm{1,1}})\right],z_e=\left[\begin{array}{c}z\\ 1\end{array}\right];$ $A=[a\left({\mathrm{\θ}}_{\mathrm{1,1}}\right)a\left({\mathrm{\θ}}_{\mathrm{1,2}}\right)...a\left({\mathrm{\θ}}_{{1,N}_2}\right)a\left({\mathrm{\θ}}_{\mathrm{2,1}}\right)...a\left({\mathrm{\θ}}_{N_1,N_2}\right)];---\left(4\right)$ θ _i，jThe element that lists for the capable j of i of matrix Θ:

Here N₁、N ₂Be the grid number that 3 n-dimensional sphere ns are divided, General N₁Desirable 10～20, N₂Desirable 5～10;The N of corresponding Θ₁×N ₂Group angle θ_1，1，θ _1，2…， θ _2，1，…，

, be 1 with element definition corresponding with the immediate angle group of DOA of the 1st user's who estimates the 1st paths among the z, all the other elements all are defined as 0; $a(\mathrm{\φ},\mathrm{\ψ})=[e^{j{\mathrm{\ω\τ}}^1}{e}^{j{\mathrm{\ω\τ}}^2}...e^{{\mathrm{j\ω\τ}}^M}{]}^T$ Guiding vector for the array of M array element; φ wherein, ψ is azimuth and the elevation angle of arrival bearing DOA, τ^m＝u(φ，ψ)r _m/c，m∈{1，2，…，M}，u(φ，ψ)＝[cosφcosψ，sinφcosψ，sinψ]， r _m＝[x _m y _m z _m] ^TBe the space coordinates of m array element, c is the light velocity, and ω is carrier angular frequencies,

And the r (n) in (3) formula is that array antenna is at the received signal vector of n sampling instant.

3, three-dimensional Beamforming Method in the smart antenna according to claim 2 is characterized in that getting d₁＝d ₂=λ/2, λ is carrier wavelength.

Images