JPH04284030A - Method for reducing convergent time of echo eraser - Google Patents

Abstract

PURPOSE:To shorten the convergent time of an echo eraser by enlarging a correction coefficient when residual echo is high and reducing the correction coefficient when the residual echo gets low. CONSTITUTION:Correlation equipments 200 and 300 are provided for calculating the level of the residual echo. Namely, residual echo 107 and an output signal 105 of a non-cyclic filter are supplied to the correlation equipment 200, the both are made correlative and an output signal alphac is generated. Namely, the correlation equipment 200 calculates an index for the size of the residual echo component corresponding to the non-cyclic filter. On the other hand, the residual echo 107 and a weighted output signal 104 of a first cyclic filter are supplied to the correlation equipment 300, the both are made correlative, and an output signal alphar is generated. Namely, the correlation equipment 300 calculates an index for the level of the residual echo component corresponding to the first cyclic filter. Therefore, in the case of arithmetic to calculate the tap coefficient of the non-cyclic filter, the outside tap coefficient of the cyclic filter and an attenuation coefficient, the correction coefficients alphac and alphar can be changed corresponding to the level of the residual echo.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for shortening the convergence time of an echo canceler that occurs in a hybrid coupling for converting a 4-wire circuit into a 2-wire circuit, and more particularly to This paper relates to a method for shortening the convergence time of an echo canceler used in digital transmission systems.

0002

2. Description of the Related Art Conventionally, a two-wire subscriber transmission system using an echo canceler has been known. Such echo cancelers are generally configured with an acyclic filter (transversal filter); A configuration method is also known in which a recursive filter is added to effectively eliminate the temporally long tail of the echo waveform. The latter echo canceler is described in Japanese Patent Laid-Open No. 63-214024.

FIG. 4 shows the configuration of this type of echo canceler. Referring to FIG. 4, 1 to 7 are shift registers;
11 to 14 are tap coefficient (weight) multipliers of the acyclic filter (transversal), 21 is an external tap coefficient multiplier of the first cyclic filter, and 22 and 23 are the first
and an attenuation coefficient multiplier of the second recursive filter, 31
and 32 are adders of the first and second recursive filters, 41 and 43 are adders, and 42 is a subtracter.

The input signal 101 to the acyclic filter is a
n and the input signal 102 of the first recursive filter is denoted an-N. The output signal 103 of the first cyclic filter is applied to the shift register 6 and the adder 43 as a weighted output signal (wn) 104 of the first cyclic filter in the external tap coefficient multiplier 21 . The output signals of the tap coefficient multipliers 11 to 14 are added by an adder 41 to become an output signal 105 of the acyclic filter, which is supplied to an adder 43. Output signal (vn) of acyclic filter 105
and the weighted output signal 104 are added by an adder 43 to form an echo replica 108. The subtracter 42 is provided with an echo waveform 106 as yn, an echo replica 108, and a residual echo 107 (indicated by en).
Output. Note that 108 is an input signal to the second recursive filter, and 109 is an output signal (pn) of the second recursive filter.

Here, the coefficients ck of the tap coefficients (weight) multipliers 11 to 14 of the acyclic (transversal) filter are
update to the nth power (where k=0.1,...,N-1), the coefficient c of the external tap coefficient multiplier 21 of the first recursive filter
The updating of I to the nth power and the updating of the coefficient r of the attenuation coefficient multipliers 22 and 23 of the first and second recursive filters to the nth power are performed by the calculations shown in equations (1) to (3), respectively. Ru.

[0006]

[0007]

[0008]

[0009] However, αc, αI, and αr are correction coefficients, each having a predetermined value.

0010

[Problem to be solved by the invention] By the way, JP-A-63
The echo canceler coefficient control method described in Japanese Patent No. 214024 has a problem in that the tap weighting coefficients take a long time to converge.

It is an object of the present invention to provide a method for reducing the convergence time of an echo canceller.

0012

According to the present invention, an echo canceler includes an acyclic filter and a cyclic filter connected in cascade to the acyclic filter, An echo replica is generated by adding the output via the external tap weight of the filter with the output of the acyclic filter, and the echo replica is subtracted from the received signal containing the echo to obtain a post-difference signal. The correction coefficient used for modifying the tap weight of the acyclic filter is adaptively controlled by the correlation value between the error signal and the output of the acyclic filter, and at the same time, the A modification coefficient used for modifying an external tap weight of the recursive filter and an attenuation coefficient in the feedback path is adaptively controlled by a correlation value between the error signal and the output of the recursive filter. shall be.

[0013]

Embodiment FIG. 1 shows an embodiment of the present invention. In this figure, blocks indicated by the same reference numbers as in FIG. 4 represent the same functional blocks. The difference between FIG. 4, which shows the configuration of a conventional echo canceller, and FIG. 1, which shows an embodiment of the present invention, is that in FIG.
It is at the point where 0 is added. Except for this point, the configuration of the echo canceller is the same.

The key point of the present invention is that the tap coefficient (weight) multiplier 1 of the above-mentioned acyclic (transversal) filter
Coefficient ck of 1 to 14 to the nth power (k=0.1,...
N-1), updating of the coefficient cI of the external tap coefficient multiplier 21 of the first recursive filter to the nth power, and attenuation coefficient multipliers 22 and 23 of the first and second recursive filters. coefficient r
In the arithmetic expressions (1) to (3) for executing the n-th power update, the correction coefficients αc, αI, and αr are adaptively changed according to the magnitude of the residual echo. That is, when the residual echo is large, the correction coefficient is increased, and when the residual echo is small, the correction coefficient is decreased, thereby shortening the convergence time of the echo canceller.

Correlators 200 and 300 in FIG. 1 are provided to obtain an index of the magnitude of the residual echo. In the figure, the correlator 200 has a residual echo (en).
)107 and the output signal of the acyclic filter (vn)1
05 is supplied. In the correlator 200, en and v
n to produce an output signal αc. The correlator 200 has a function of obtaining an index of the magnitude of the residual echo component corresponding to the acyclic filter. On the other hand, the correlator 300 is supplied with a residual echo (en) 107 and a weighted output signal (wn) 104 of the first recursive filter. In the correlator 300, en and wn are correlated to generate an output signal αI or αr. The correlator 300 has a function of obtaining an index of the magnitude of the residual echo component corresponding to the first recursive filter.

FIG. 2 shows an example of the circuit configuration of correlator 200 or 300. The figure shows a correlator 300 as an object. In the figure, 501, 502, and 503 are multipliers, 504 is an adder, and 505 is a shift register, respectively. The residual echo (en) and the weighted output signal (wn) of the first recursive filter are supplied to a multiplier 501, and the multiplication result is input to a multiplier 502, multiplied by Υ (Υ>0) and sent to an adder. 504. Adder 5
04, a closed loop circuit consisting of a shift register 505 and a multiplier 503 constitutes a first-order integrator, and the multiplier 503 multiplies by a constant (1-β). However, 0<β
<1. The output of the adder 504 is αr to the nth power or αI to the nth power, and is used as the correction coefficients αr and αI in equations (2) and (3). The circuit shown in FIG. 2 can be expressed as an arithmetic expression as follows.

[0017]

However, although the above equation focuses on αr, αI can also be expressed in the same way. Also, αc
can also be expressed as an arithmetic expression similar to equation (4). At this time, in FIG. 2, if the signal wn supplied as an input to the multiplier 501 is replaced with vn, the output of the adder 504 is αe raised to the nth power. In equation (1), αc to the nth power is used as αc.

As explained above, in the calculation formulas (1) to (3) for determining the tap coefficient of the acyclic filter, the external tap coefficient of the recursive filter, and the attenuation coefficient, the correction coefficient α
It becomes possible to adaptively change c, αI, and αr according to the magnitude of the residual echo, and it is possible to shorten the convergence time of the echo canceller.

[0020]

The characteristics of the echo canceller according to the present invention described above are shown below. As an example, consider a case where the echo waveform has a negative exponential response corresponding to a low cutoff frequency of 1/200 of the data transmission rate, and the main part of the echo waveform is N = 32.
It is assumed that the echo can be canceled by a tap acyclic filter, and equation (5) is used for the "tail" of the echo.

[0021]

[0022] However, hi (i=0.1,..., N, N+
1,...) represent the impulse response of the echo. Also, hN
=-0.0118 rO =exp(-2π/200)=0.96907. By computer simulation, the residual echo average value (Equation (6))

[0023]

FIG. 3 shows the results of an investigation into how . In this simulation, an is +1
Or -1 random series, and the initial value of the coefficient is the formula (
7) to (9).

[0025]

[0026]

[0027]

The average value ε of the residual echoes is expressed in decibels. As is clear from FIG. 3, the convergence time of the echo canceller is significantly shortened compared to the conventional convergence method in which the correction coefficient is fixed, and it can be seen that the method of the present invention is effective.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an echo canceler according to the present invention;

[Figure 2]
A configuration diagram of the correlator in FIG. 1,

[Fig. 3] A diagram showing an example of simulation results of this example,

FIG. 4 is a block diagram of a conventional echo canceler.

[Explanation of symbols]

1, 2, 3, 4, 5, 6, 7, 505 Shift register 11, 12, 13, 14 Tap coefficient multiplier 21
External tap coefficient multiplier 23 Attenuation coefficient multiplier 501, 502, 503 Multiplier 31, 32, 4
1, 43, 504 Adder 43 Subtractor 200, 300 Correlator

Claims (1)

[Claims] 1. A method for shortening the convergence time of an echo canceller used in two-wire bidirectional digital data transmission, wherein the echo canceller includes an acyclic filter and a cyclic filter connected in cascade to the acyclic filter. filter, the output of the recursive filter via external tap weights is added to the output of the acyclic filter to generate an echo replica, and the echo replica is generated from the received signal containing the echo. The configuration is such that an error signal is obtained by subtracting echo replicas, and a correction coefficient used for modifying the tap weight of the acyclic filter is correlated with the error signal and the output of the acyclic filter. At the same time, each correction coefficient used to modify the external tap coefficient of the recursive filter and the attenuation coefficient in the return path is combined with the error signal and the output of the recursive filter. A method for shortening the convergence time of an echo canceller, characterized by adaptively controlling the correlation value of the echo canceller.