introduction
At present, due to the rapid development of mobile communications, the number of mobile users has advanced, which will lead to limited spectrum resources being utilized by "unlimited", and the contradiction is very sharp. How to effectively utilize spectrum resources is a problem that is difficult to avoid in the development of mobile communication, and the emergence of intelligent antennas brings life to mobile communications. It can effectively utilize spectral resources, improve system capacity, is a key technique that is essential in future mobile communications. Adaptive beamforming algorithm is the core of intelligent antenna research. In the CDMA system, different users have different PN codes, can different PN codes to implement beam-shaped algorithms? Rong. Z et al. Is based on this idea, proposing the least Square Slimness Reagent, LS-DRMTA, Least-Squares Despread Respread MultiTargetArray and the Minimum Slim Sixth Separation Expand Multi-Target Constant System Array Algorithm ( LS-DRMTCMA, Least-Squares Despread Respread MultiTargetConstant Modulus Array. These two algorithms have a lot of advantages, and the price has increased computational complexity. On the basis of the literature, the DR-LMS algorithm is proposed. This paper first introduces the LS-DRMTCM algorithm, and then introduces the DR-LMS algorithm in detail, and finally puts forward a new transformation of the algorithm based on the idea of the algorithm improvement in the literature. The new algorithm has made MATLAB simulation.
1, signal model
A DS-CDMA system having k users, the receiving end is a uniform linear array having a M unit. Assume that the power of the kth user is PK, DOA is θk, the array response matrix is
2 Algorithm
2.1 LS-DRMTCMA and DR-LMS algorithms
In the CDMA system, the user's PN code is known. At the receiving end, the spread spectrum signal generated by the PN code of the user i is recorded as a CI (T), and the Ci (T) delay τi is delayed and the received signal is correlated, and the decision is performed after processing, and the nth information bit decision. The result is bin, if the decision is correct, that is, there is bin = bin, the Ci (T) delay is retired to the BIN, and the user I declares the redemption signal RIPN, it and the input signal Yi ( T) hard-limited signal RCIM performs weighting and summing to construct the transmit waveform Ri (t) of user I at time [N 1) TB, NTB], which is the LS-DRMTCM algorithm, and its cost function can be expressed as:
Here μ is the step size, the convergence speed of the algorithm. The initial value is set to
2.2 Improvement Algorithm
This paper modifies the DR-LMS algorithm to a new algorithm. Since the reference signal RI (M) of the algorithm is the despreading re-expanded, the algorithm is still blindly blind. The algorithm has continued the advantages of using the PN code characteristics in the above two algorithms, and the improved algorithm is:
In the algorithm, μ (m) is the iterative step size of the algorithm, the convergence speed of the control algorithm, can be seen from the step size adjustment principle, the step length should be larger in the initial stage of the algorithm iteration, in order to obtain a large convergence speed, but In the convergence stage, regardless of the measurement noise, it should be smaller steps to achieve lower steady state disorders. EIL (M) is the generated error signal; μOPT is the fastest time when the algorithm is constant; α is the adjustment factor, the value range is set to 0.1 "α" R, R is the algorithm convergence The ratio of the state average error and noise variance; β is a smoothing factor, the value range is O "β" L; σ2n is a variance of measuring noise N (N).
2.3 Effect of step length factor on algorithm
In the formula (13), the H is in the LS-DRMTCM algorithm, which is the size of the data sample, which is the average value of the sample data block. According to the literature, in any iterative phase, the mean square value of the output error must be greater than the variance of the measurement noise, and therefore, it will be known that it will be established.
2.4 Impact of α and β on the performance of algorithm
Different selection of parameters α and β can affect the convergence speed and steady distortion of the algorithm. Here, the value of the parameter A is determined by the method of the test, first gives an alpha value, such as α = 0.5, so it can obtain a learning curve, then gradually change the α value to obtain a group of learning curves, Select a curve in which the convergence effect is best to determine the α value. It is understood from the formula (13) that when the alpha is faster, but the smoothness is poor, and when α is large, the convergence speed of the algorithm is slow. Therefore, the appropriate α value can be selected, enabling algorithms to obtain both faster convergence speeds and relatively low steady state disorders. The effect of β is to smooth the steps of the steps. If the adjacent point of the curve is large, the large beta value should be selected, but in turn, the smaller value should be selected. In this way, the fluctuations of EIL (M) have a large fluctuations in the step length factor μ (m) to achieve better convergence performance.
3 simulation results
It is assumed that in the Gauss white noise channel, the base station antenna is an equal interval linear array of 8 cells; the interval between the cells is half a carrier wavelength, the signal-to-noise ratio is 15 dB; the signal noise ratio is 10dB; the spreading factor is 3L; the desired signal The angle of incidence is 30 °, the interference direction is one 50 °, and the number of iterations is 1000 times.
3.1 Convergence Performance
The DR-LMS algorithm and the new algorithm all samples only calculate one weighted vector in a bit cycle. Fig. 1 is the convergence curve of the two algorithms, and it is set to μ = O in the DR-LMS algorithm. 000045, in the new algorithm, α = 0.8, β = 0.2, and the starting step μOOPT = 0.000045. As can be seen from the convergence of the two algorithms: Under the same conditions, the DR-LMS algorithm can converge in the iteration of about 500 times, and the algorithm proposed in this paper can converge, the convergence speed is obvious. It is better than the algorithm mentioned in the literature.
3.2 Borenap
As can be seen from Figures 3 and 4, the DR-LMS algorithm and the algorithms proposed in the literature can be well formed in the desired direction, and the suppression of the interference direction signal is also obvious.
3.3 Comparison of complexity of algorithm
In the literature, the complexity of the LS-DRMTCM algorithm is h (2 m2 + m), where h is the size of the sampling data block, and m is the number of array elements of the array antenna. In the literature, the complexity of the DR-LMS algorithm is 2HM, and the algorithm proposed in this article has joined the DR-LMS algorithm, but this does not increase the complexity of the algorithm, and the algorithm is not included in the algorithm. Any exponential operation, the calculation is simple, only a minimum multiplication operation, so the computational complexity is low, and the amount of DR-LMS algorithm is calculated is substantially.
4 small knot
In this paper, the intelligent antenna technology in mobile communication is the research background, and the blind adaptive beamforming algorithm based on code filtering is studied. The literature proposes the DR-LMS algorithm based on RONG et al., This article is in this idea. The DR-LMS algorithm has been further improved under the guidance, and the transformation algorithm is introduced. By simulation comparison, the algorithm mentioned in the tracking performance and the algorithm mentioned in the literature is comparable, but there is significant increase in convergence performance.
Editor in charge: GT, read full text
Our other product:
