JPS583414A - Impulse responsive automatic equalizer - Google Patents

Abstract

PURPOSE:To reduce the arithmetic quantity which is required for training and to save the hardware, by compensating a tap coefficient so that the output of the 2nd filters symmetrical to each other centering on the center tap become an impulse. CONSTITUTION:In a training time, a complex conjugate series is produced at a coefficient generating part 2 from the impulse series which is reproduced at an impulse detecting part 1 and then applied to an input of a correlator 3. At the same time, the impulse series is applied to the other input of the correlator 3. Thus an own correlation value series is obtained. A matrix equation is solved to the output of the correlator 3 at an initial tap coefficient arithmetic part 4. At the same time, the output of the correlator 3 is fed to an amplitude normalizing part 4-1 to normalize the amplitude, and then the matrix equations are solved successively at a solution arithmetic part 4-2. A convolutional operation is carried out at a convoluting part 5 for a series of resolutions obtained through the part 4 and the tap coefficient of the correlator 3. Thus the tap coefficient is obtained for an automatic equalizer 7. Then the initialization is carried out to a tap coefficient compensating part 6.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides training by impulse,
The present invention relates to an impulse response automatic equalizer that sets a tap coefficient sequence in a transversal filter.

An automatic equalizer is essential in data modems, etc., and is used to compensate for waveform distortion of the received signal by estimating the distortion in the transmission path based on the received signal and automatically setting compensation characteristics. Generally, a transversal filter is used to automatically adjust the tap coefficient sequence to obtain a desired compensation characteristic.

In such an automatic equalizer, before performing actual data transmission, it is necessary to perform an operation (training) to set equalizer characteristics by sending a specific signal through a transmission path in advance. There are two methods for training an automatic equalizer: one using a pseudorandom code sequence and the other using impulses.The latter method is superior because it takes less time for the automatic equalizer to converge. ing.

There are two methods of training Il-formed i using Inhalsler: a method of directly solving the determinant created by the cross-correlation between the impulse response series of the transmission path and the error signal, and finding the tap coefficient sequence; After Fourier transform and replacing it with frequency domain and flattening it,
There is a method of determining the Princess Tube coefficient sequence by performing inverse Fourier transform, but either method requires a huge amount of calculations and inevitably increases the circuit scale.

In contrast, the same applicant filed a patent application in 1984-126.
No. 487 proposed a method in which the values of the input series are given as they are as the initial values of the tap coefficients, and then the °C tap coefficients are corrected by the method of least squares so that the output impulse response becomes 0 except for the median value. There is. This method can perform automatic equalization using relatively simple means and can be realized on a normal hardware scale, but if you try to widen the tail of the impulse response, you will need to use rehardware, which still requires a large amount of calculations. could not be avoided from increasing.

The present invention attempts to eliminate the drawbacks of the prior art, and its purpose is to further reduce the amount of computation required for running, thus saving 8-domain space. The object of the present invention is to provide an automatic equalizer that can perform the following steps.

The impulse response automatic equalizer of the present invention includes a first transparsal filter that uses the impulse response sequence as its tap coefficients, and a second variable tap transpersal filter that is symmetrical about the center tap, The tap coefficient of the second filter is corrected so that the output of the second filter becomes an impulse, and by making the number of taps of the second filter short and symmetrical, the tap coefficient This reduces the amount of calculation required for the correction of , thereby saving 8-domain air. Note that the function of the first filter is equivalent to performing a correlation operation on the input signal sequence.

Hereinafter, the present invention will be described in detail with reference to Examples.

The impulse response automatic equalizer of the present invention includes two transparsal filters, and the output sequence of the first transparsal filter is an autocorrelation sequence of the input sequence since the input sequence is directly placed in the tap sequence sequence. Therefore, it is symmetrical about the median value. By passing this output series through a symmetrical second variable tap transversal filter, an equalized impulse is obtained as an output.

Now, the input sequence of the second filter is (X−N @X−
w+1e"・*X-1e xo'+X1*"・
@XI) and the second filter 9f)97 coefficient as (
C-M e ”' t CO* CI'* ・・・v
CM), this tap coefficient is the solution of the following equation (C:)
As long as it is requested.

Since the diagonal elements of (X)ma) 9 y i are always on the dominant balance, the above equation can be easily solved by Gachseidel's sequential solution method (Riley engineering mathematics).

It is also possible to use the symmetry to make the order a gutter.

FIG. 1 shows the configuration of an embodiment of an automatic impulse response equalizer of the present invention. In the figure, 1 is an impulse detection section, 2 is a coefficient generation section, and 3 is a correlator. 4 is an initial tap coefficient calculating section, and amplitude normalizing section 4-1. It consists of a sequential solution calculation section 4-2. 5 is the convolution part, 6
is a tap coefficient correction unit, 7 is an automatic equalizer, 8 is a buffer,
9 is a determination section, and 10 is an error generation section.

In FIG. 1, an incoming signal sequence input from a transmission line having a training signal containing an impulse added to the beginning of data is input to an impulse detection unit 1, and the impulse detection unit 1 calculates the impulse response of the transmission line from this. Play as a waveform. This process is based on the patent application filed in 1973 by the same applicant.
As shown in No. 5-172091, when a signal sequence containing impulses is modulated with a periodic signal, the waveform can be directly reproduced by calculating the sum or difference of the signal sequence. Alternatively, if the impulse-containing signal is generally some defined Kundam sequence, it can be determined by correlating the incoming signal with the random sequence.

In this case, in the case of a transmission format such as an orthogonal amplitude modulation method, the incoming signal sequence is handled as a complex number.

The coefficient generating section 2 extracts several samples of the impulse waveform required for equalization from the impulse sequence reproduced by the impulse detecting section 1, creates a complex conjugate sequence thereof, and supplies the complex conjugate sequence to one input of the correlator 6. The reproduced impulse sequence is input as is to the other input of the correlator 6, so that the output of the first 29 filter, that is, the correlator 3, is the output of the reproduced impulse sequence. A correlation value series is obtained.

〉′□. 5 t) i @ Q engineering□2,1 yo, ga26.6
゜It can be expected that the diagonal elements are larger than the other elements. Therefore, by taking advantage of this feature, the determinant can be found relatively easily.

The initial tap coefficient calculating section 4 executes a method of solving the determinant for the output of the correlator 3. The output of the correlator 3 is input to the amplitude normalization section 4-1, and the amplitude of the seven is normalized so that the diagonal elements of the matrix are set to 1.

This is to make it easier to solve the matrix, and the amplitude-normalized signal is sequentially input to the solution calculation section 4-2, where the determinant solution calculation is executed. Here, the multidimensional simultaneous equations represented by the above-mentioned equation (1) are sequentially solved by, for example, the Gak-Diedel solution method.

The convolution unit 5 performs a convolution operation on the series of sequential solutions obtained from the initial tap coefficient calculation unit 4 and the tap coefficients in the correlation a3 to obtain tap coefficients for the automatic equalization amount 7. At the training stage, the switch 8W4 is open and the switch 8W3 is closed, so the tap coefficients determined by the convolution section 5 are initialized in the tap coefficient correction section 6.

The tap coefficient correction section 6 determines the tap coefficients of the automatic equalizer 7, and thereby the characteristics of the automatic equalizer 7 are set. The series of processes described above are processed at high speed except for the delay time required to input the impulses, so the optimal tap coefficients of the automatic equalizer can be found extremely quickly.

The training is completed by initializing the tap coefficients in the tap coefficient correction section 6, and the switch 8W5 is opened and the switch 8W4 is closed.

Data is input from the transmission path following the impulse, and the data is stored in the buffer 8. At the end of the training, the data in the buffer 8 is input to the automatic equalizer a7 and equalized, and the equalized data is input to the determination section 9 where the sign is determined and an output is produced. The output of the determining section 9 is inputted to the signal generating section 10, where it is compared with a reference signal and a minute F-component is extracted. The error component extracted by the error generation section 9 is fed back to the tap coefficient correction section 6 via the switch 8W4. The tap coefficient correction unit 6 adaptively corrects the tap coefficients of the automatic equalizer according to the feedback error components using the conventional maximum likelihood gradient method, so that the characteristics of the automatic equalizer are always maintained. Keep it in optimal condition.

FIG. 2 shows the structure of another embodiment of the automatic impulse response equalization amount of the present invention. In the figure, 3A indicates a fixed equalization amount, and the correlator in the case of FIG. 1 is constantly provided as a fixed equalizer. Moreover, the convolution part in the case of FIG. 1 is missing. Note that the same parts as in FIG. 1 are indicated by the same numbers.

In FIG. 2, in the training state, the swings j8W1 and 8W2 are connected to the contact a side, and the complex impulse sequence detected by the impulse detection unit 1 from the incoming signal sequence passes through the coefficient generation unit 2 and is converted into a complex conjugate. A sequence is created and input to the fixed equalizer 6A, and
The signal is input as is to the fixed equalizer 6A. As a result, the fixed equalizer 6A generates an autocorrelation value series of the impulse series.

The autocorrelation value series is first
The process is performed in the same way as in the figure to obtain sequential solutions. The output of the initial tap coefficient calculation section 4 is directly input to the tap coefficient correction section 6 as the tap coefficient of the automatic equalizer 7 via the switch 8W5 without being convolved with the tap coefficient in the fixed equalizer 3A. Prized.

When the initialization of the tap coefficient in the tap coefficient correction section 6 is completed, the switches 8W1 and 8W2 are all switched to the contact side, and the switch SW3 is opened and the switch SW4 is switched to the contact side.
closes. As a result, the data stored in the buffer 2 is input to the fixed equalization amount 3A via the switch Sw1, producing a somewhat equalized output. This output is input to the automatic equalization amount 7 via the switch 8W2, and is equalized using the tap coefficients already set in the tap coefficient correction section 6. The output of the automatic equalizer 7 is judged by a judging section 9 to produce an output. The error component is extracted from the output by the error generating section 1, and the extracted error component is fed back to the tap coefficient correction section through the switch SW4 to perform adaptive correction, as in the case of Fig. 1. .

In the embodiment shown in FIG. 2, the correlator is constantly provided as a fixed equalizer, and an automatic equalizer is placed at the subsequent stage, and the folding operation is omitted. The amount of computation is further reduced.

As explained above ;) K. With the automatic impulse response equalization amount of the present invention, there is no need for procedures such as Fourier transform to determine the tap coefficients, and the transversal filter is controlled in two parts. By making the number of taps of the second filter short and symmetrical, the amount of calculation for correcting the tap coefficients is reduced, and hardware is saved. Furthermore, since the number of taps is short, convergence is quick, and therefore performance as an automatic equalizer can be improved.

[Brief explanation of drawings]

FIGS. 1 and 2 are block diagrams each showing the configuration of an embodiment of an automatic impulse response equalizer of the present invention. DESCRIPTION OF SYMBOLS 1... Impulse detection part, 2... Coefficient generation part, 6.
... Correlator, 5A... Fixed equalizer, 4... Initial tap coefficient calculation section, 4-1... Amplitude normalization section, 4-2 River sequential solution calculation section, 5... Convolution section, 6... tap coefficient correction section, 7... automatic equalizer, 8... buffer,
9...Determination Department, 1゜...Error Occurrence Department Patent Applicant Fujitsu Limited Representative Patent Attorney Gobe Tamamushi and 3 others

Claims (1)

[Claims] In an automatic equalizer that automatically sets tap coefficient values in a transpertil filter to equalize transmission path distortion,
A first transversal filter whose tap system is an impulse response sequence reproduced from a training signal containing impulses from a transmission path, and a second variable tap transversal filter symmetrical about a center tap. An impulse response automatic equalizer, characterized in that equalization is performed by correcting tap coefficients of the second transpersal filter so that the output of the second transpersal filter becomes an impulse.