KR0152028B1 - Self-recovering equalizer and initializing method thereof by spectrum prediction - Google Patents

Abstract

The present invention relates to a magnetic recovery equalizer using spectral prediction and a method of initializing the equalizer using a simplified method of spectral prediction when initializing coefficients of the magnetic recovery equalizer. The present invention includes a finite shock response filter for delaying a received signal, multiplying the coefficient values, and adding all the multiplied values. The initial coefficient calculation unit for applying an initial coefficient value to the finite shock response filter receives the received coefficient from the received signal. The process of calculating a new eigenvector by calculating the autocorrelation matrix, reducing the amount of computation using the Levinson-Dubin algorithm, and calculating the new eigenvector by calculating the new eigenvector using the matrix and the initial eigenvector Repeatedly, the final calculated eigenvector is set as an initial coefficient value. At this time, the communication channel does not change rapidly and the time required for initializing the equalizer can be reduced by using the general channel characteristic that the autocorrelation matrix of the received signal has a hermit topless characteristic.

Description

Magnetic Recovery Equalizer Using Spectral Prediction and Its Initialization Method

1 is a block diagram of a magnetic recovery equalizer using the spectral prediction of the present invention.

* Explanation of symbols for main parts of the drawings

10: finite impact response filter 20: initial coefficient calculation unit

30: equalization coefficient calculation unit T1-Tn: delay element

M0-Mn: Multiplier A1-An: Adder

The present invention relates to the initialization of a channel equalizer and the improvement of convergence speed. In particular, the present invention simplifies the spectral prediction technique when initializing the coefficients of a Blind equalizer in which a training sequence is not transmitted. The present invention relates to a magnetic recovery equalizer using spectral prediction and a method of initializing the same which enable the equalizer to converge at a high speed using the method.

In general, a signal transmitted in a digital communication system passes through a channel and is distorted due to non-ideal characteristics of the transmission channel, that is, noise, nonlinear filtering, multipath, and Doppler shift, or inter symbol interference. ) Is generated so that the received signal is different from the original signal. The equalization technique means the restoration of the distorted signal to the original signal while passing through the channel. If the channel distortion is not compensated through the equalizer, the symbol error rate (SER) becomes very high and is decoded into the wrong data.

In the case where the training sequence is periodically transmitted among the transmitted signals, the equalization process can be easily performed by means of a mean square error (MSE) method. However, if the training sequence cannot be sent for various reasons, it is necessary to calculate and update the coefficients of the equalizer using only the received signal without knowing which signal was transmitted. This case is called self-recovery equalization or blind equalization. In this case, it takes considerable time to converge the channel equalizer, and the performance of the equalizer is lower than that of the training sequence. Deteriorates a lot.

Known self-healing equalization algorithms for digital communication to date include Goddard Algorithm, Decision Directed Algorithm, Stop-and-Go Algorithm. Although well compensated for, the convergence is slow and the error variation after convergence is relatively severe. Therefore, in order to solve this disadvantage, the distortion of the channel is compensated to some extent using the Goddard algorithm, and then the steady state mean square error is minimized by using the DD algorithm.

However, the DD algorithm determines that the transmittable signal having the smallest distance error between the received signal and all transmittable signals is the signal transmitted from the original transmitter, and obtains an error vector by the difference between the received signal and the determined transmittable signal. Update the coefficient of the group. At this time, if there is a little noise, it can be easily converged by obtaining the same error vector as the actual error. However, when the distortion is severe, the received signal is located at a completely different position beyond the demarcation boundary. Can't. Therefore, the DD algorithm can be used only after the channel distortion is removed to a certain level. In order to compensate for this, the SG algorithm was developed instead of the DD algorithm. SG algorithm is a kind of DD algorithm, and has a characteristic of greatly reducing the decision error that can occur in DD algorithm.

In addition to the above Goddard algorithm, another initialization method for calculating the initial coefficient of the magnetic recovery equalizer is a method of predicting the characteristics of the channel using the spectral prediction technique. This initialization method allows the magnetic recovery equalizer to converge much faster than the initialization method using Goddard's algorithm. However, this method has a disadvantage in that a considerable amount of calculation is required, so it takes a lot of time to calculate the initial coefficient of the equalizer, and the calculation method is also quite complicated.

SUMMARY OF THE INVENTION The present invention has been made to solve the conventional problems as described above, and an object of the present invention is to initialize the coefficients of the magnetic recovery equalizer by using a channel prediction method that simplifies the spectral prediction technique, thereby rapidly converging the equalizer and diverging the equalizer. The present invention provides a magnetic recovery equalizer using spectral prediction and its initialization method to reduce a large error performance.

Another object of the present invention is to provide a self-healing equalizer using spectral prediction and its initialization to simplify the channel prediction method using the general characteristics appearing in most channels, thereby reducing the amount of computation and the time it takes to calculate the initial coefficient of the equalizer. To provide a method.

The magnetic recovery equalizer using the spectral prediction according to the present invention for achieving the above objects includes a plurality of delay elements for delaying a received signal at a predetermined time interval, and coefficient values calculated according to channel characteristics in the received and delayed signals. And a finite shock response filter including a plurality of multipliers for continuously multiplying and a plurality of adders for adding all of the output signals of the multiplier and outputting a signal having a linear combination of input data and past values thereof. The initial coefficient calculator which applies an initial coefficient value to the finite shock response filter calculates an autocorrelation matrix of a received signal and then stores an eigen vector corresponding to the minimum eigen value of the matrix. The algorithm is calculated using the (Levinson-Durbin) algorithm and this value is output as the initial coefficient value of the finite shock response filter. After the initial coefficient value is determined, the output signal of the finite impact response filter is applied to the equalization coefficient calculating unit, and the equalizing coefficient calculating unit continuously recalculates the coefficient values using the existing equalization algorithm and outputs it to the finite impact response filter.

In addition, the initialization method of the magnetic recovery equalizer using the spectral prediction of the present invention for achieving the above objects, the method comprising the steps of obtaining an autocorrelation matrix from the received signal, determining an arbitrary initial eigenvector, and the autocorrelation matrix And calculating a new eigenvector by the Levinson-Derbin algorithm using the initial eigenvector, and replacing the new eigenvector with a previous eigenvector to calculate a new eigenvector and repeating this process a predetermined number of times. Are sequentially performed to set the final calculated eigenvector to an initial coefficient value. At this time, the communication channel does not change its characteristics rapidly, and the time taken to initialize the equalizer by using the general channel characteristics that the autocorrelation matrix of the received signal has the symmetrical shape of the symmetry around the longest diagonal line. And reduce the amount of computation.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the attached FIG. 1.

1 is a block diagram of a magnetic recovery equalizer using spectral prediction proposed in the present invention. The magnetic recovery equalizer is composed of a finite impulse response filter (10), and the input signal of the magnetic recovery equalizer is demodulated through a channel and then input in the form of a digital signal. The magnetic recovery equalizer includes a plurality of delay elements T1 -Tn for delaying an input digital signal at predetermined time intervals, and the delay elements T1 -Tn are connected in series with each other. A plurality of multipliers M0-Mn are connected to an input terminal of the magnetic recovery equalizer and an output terminal of delay elements T1 -Tn to continuously multiply the input and delayed digital signals with coefficient values calculated for channel characteristics. . The output signals of the multipliers M0-Mn are added to the adders A1-An, and the signals are added together in the adder An so that a signal having a linear combination of input data and the past is equalized. It is output to 30. When the output signal of the adder An is viewed on the time axis, the shock response is represented in the form of a finite length.

At this time, the digital signal input to the magnetic recovery equalizer is applied to the initial coefficient calculating unit 20, where an autocorrelation matrix of the input signal is calculated and then an eigenvector corresponding to the minimum eigenvalue of the matrix is calculated. This value is applied to the multiplier M0-Mn as the initial coefficient value of the magnetic recovery equalizer. After initializing the magnetic recovery equalizer, the multiplier M0-Mn multiplies the digital signal by the coefficient value applied from the equalization coefficient calculator 30 to compensate for the distorted signal while passing through the channel. The equalization coefficient calculator 30 continuously adjusts the coefficient value of the magnetic recovery equalizer by using an existing equalization algorithm, for example, the DD algorithm, the SGA algorithm, the MSGA algorithm, and the like to quickly adjust the coefficient value to the channel characteristics. Allow convergence

The present invention proposes a method for calculating an optimal initial coefficient value in a magnetic recovery equalizer configured as described above, and shows a method of greatly simplifying a spectral prediction method by obtaining an eigenvector using general characteristics appearing in most channels. . In this case, the general communication channel has the following characteristics. First, a communication channel is a weakly stationary random processor whose characteristics do not change drastically. Second, a signal sequence represented by a white random processor, that is, samples having no correlation between neighboring samples, communicates. The autocorrelation matrix of the signal received through the channel is Hermitian Toeplilz. In more detail, the autocorrelation matrix satisfies the formula (R ^* ) ^T = R and has the same value in each diagonal direction. Where * represents a complex conjugate, T represents a transposition, and R is a complex square matrix. For example, R is the matrix

By obtaining the autocorrelation matrix of the received signal and obtaining the eigenvector corresponding to its minimum eigenvalue and using it as the initial coefficient of the self-recovery equalizer, it is possible to converge the equalizer much faster than the method using the Goddard algorithm. Can also be reduced a lot. To calculate the eigenvectors corresponding to the minimum eigenvalues, a considerable amount of computation is required. However, the above two channel characteristics can be used to significantly reduce the computations.

The coefficient initialization algorithm of the magnetic recovery equalizer of the present invention is as follows. First, the autocorrelation matrix R is obtained from the received signal by the following equation.

Where E [] represents an ensemble average, C is a vector of the received signal, * is a complex conjugate, and T is a transposition. Second, an arbitrary initial eigenvector [V (0)] is determined. For example, V (0) = [1, 1,... , 1]. In this case, the matrix order of the eigenvectors is equal to the order of the magnetic recovery equalizer. Third, a new eigenvector [V (k + 1)] is obtained by solving the following function.

Here, V (k + 1) can be easily calculated using the Levinson-Derbin algorithm. This is because R has a hermit topless characteristic. Fourth, V (k) = V (k + 1). Fifth, repeat the third and fourth processes m times. Here, the greater the number of repetitions (m), the better the channel equalization performance of the magnetic recovery equalizer, but the proper number should be determined in consideration of the calculation time. Sixth, the finally calculated eigenvector [V (k)] is used as the initial coefficient of the magnetic recovery equalizer. After initializing the magnetic recovery equalizer using this coefficient initialization algorithm, if the coefficients of the magnetic recovery equalizer are continuously readjusted to the most suitable values for the channel characteristics by using the existing equalization algorithm, the magnetic recovery equalizer can be quickly restored. It can converge stably.

As described above, the present invention initializes the coefficients of the self-healing equalizer by a greatly simplified spectral prediction method using general channel characteristics, thereby reducing the time and calculation amount required to initialize the equalizer, thereby preventing the equalizer from diverging and providing a fast time. There is an effect that can be converged stably within.

Claims (7)

A self recovery equalizer for compensating for distortion and interference of a signal received through a communication channel, comprising: obtaining an autocorrelation matrix from a received signal; Determining any initial eigenvectors; Calculating a new eigenvector by the Levinson-Dervin algorithm using the autocorrelation matrix and the initial eigenvector; Replacing the new eigenvector with a previous eigenvector to calculate a new eigenvector and repeating this process a predetermined number of times; And setting the final calculated eigenvector as an initial coefficient of the magnetic recovery equalizer. 2. The method of claim 1, wherein the communication channel is a random processor whose characteristics do not change drastically. The spectral prediction of claim 2, wherein the autocorrelation matrix of a signal received through uncorrelated signal strings through a communication channel has a helical topless characteristic that is symmetrical about a longest diagonal line. Initialization method of magnetic recovery equalizer using. 4. The method of claim 3, wherein the autocorrelation matrix (R) is obtained by the following equation. Where E [] represents the ensemble mean, C is the vector of the received signal, * is the complex conjugate, and T represents the transpose value. 5. The method of claim 4, wherein the new eigenvector [V (k + 1)] is calculated by solving the following function. Where V (k) is the previous eigenvector. 6. The method of claim 5, wherein the matrix order of the eigenvectors is set equal to the number of coefficients of the magnetic recovery equalizer. A magnetic recovery equalizer for compensating for distortion and interference of a signal received through a communication channel in which channel characteristics do not change drastically, comprising: a plurality of delay elements connected in series with each other to delay the received signals at predetermined time intervals; A plurality of multipliers for continuously multiplying the received and delayed signals with coefficient values calculated according to channel characteristics; A plurality of adders for adding all of the output signals of the multiplier to output a signal having a linear combination of input data and past values thereof; An initial coefficient calculator for calculating an autocorrelation matrix of the received signal and calculating an eigenvector corresponding to the minimum eigenvalue of the matrix using a Levinson-Derbin algorithm and outputting this value to the multiplier as an initial coefficient value; And an equalization coefficient calculator which receives an output signal of the adder, calculates a coefficient value using an equalization algorithm, and continuously readjusts the coefficient value and outputs it to a multiplier.