JPS6173433A - Removing method of echo - Google Patents

Abstract

PURPOSE:To simplify control, to reduce the scale of a circuit and to shorten a converging time by adding a circuit consisting of a subtractor and a delay element for applying the delay of Tsec (one bit width), an interpolation filter and a circuit consisting of polarity detectors, a correlator and a multiplier to a conventional circuit. CONSTITUTION:A value obtained by subtracting a value obtained before Tsec from a current value appears in the output of a subtractor 16 by adding the subtractor 16 and the delay element 17. Consequently, a condition that the probability to obtaion the polarity of a residual echo component in a difference signal to be the output of a subtractor 10 precisely is not zero is obtained by a polarity detector 12 and the adaptability of an adaptive digital filter 8 is guaranteed, so that control is simplified. The correlated value of the outputs of polarity detectors 23, 19 is calculated by a correlator 20 and multiplied by 2alpha (alpha is a constant) in a multiplier 21. The polarity of said difference signal appears in the output of the detector 23 and the polarity of an each replica appears in the output of the detector 19. Therefore, the output of the correlator 20 is changed in accordance with the size of the residual echo. Thereby, the output concerned is scaled by multiplying it by 2alpha in the multiplier and used as step size, the polarity of the output of the detector 12 is applied to said step size and the added value is fed back to the filter 8, so that the converging time is sharply shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a rounding and echo cancellation method for realizing two-wire bidirectional digital transmission.

(Problems with the prior art) Echo Gear/Cera is known as a well-known technology for realizing 2-wire bidirectional digital transmission using paired wires.・And signal glossing (IEEEETRAN8ACTION8ON
ACOUSTIC8゜5PEECH, AND 5I
GNAL P几0CESS-ING) Volume 27, No. 6, 1
979, pp. 768-781). Echo canceller, generate a pseudo echo (echo replica) corresponding to the transmitted data sequence using an adaptive filter with a tap coefficient equal to the length of the echo noise response (K), 2-wire/4 The line conversion circuit operates to suppress echoes leaking from the transmitting circuit to the receiving circuit. At this time, each tap coefficient of the adaptive filter is successively corrected by correlating the difference signal obtained by subtracting the echo replica from the mixed signal in which the echo and the received signal are mixed and the transmitted data. The coefficient modification of such an adaptive filter, that is, the convergence algorithm of the echo canceller, is described in the above-mentioned reference), and a typical example is the stochastic iteration algorithm (stoch
hastic 1teration algorithm
m) and sign.

algorithm is known.

In order to realize two-wire bidirectional digital transmission, it is necessary to implement LSI, and it is desirable to use a method that can apply digital device technology, which has recently made significant technological progress. At this time, if you try to use a digital filter as the adaptive filter mentioned above, the analog/digital (A/D)
D) converter and digital/analog C)/A
＞A converter is required. Of these, the number of bits required for the D/A converter is a fixed number tb from the system requirements, for example, about 12 bits in application to a subscriber line of a public communication network. On the other hand, the required number of bits for the A/D converter depends not only on the system conditions but also on the convergence algorithm of the echo canceller described above. For example, when applied to a subscriber line of a public communication network, the Stochastic Iteration Algorithm requires about 8 bits, whereas the Sign Algorithm requires only 1 bit. However, in the sine algorithm, the tap coefficients of the adaptive filter are modified according to the polarity of the difference signal, so if the polarity of the residual echo contained in the difference signal does not match the polarity of the difference signal, The problem arises that adaptive behavior becomes impossible. For example, when a binary code such as a bi7-factor code is used as a transmission line code, the residual echo (difference between the echo and echo replica) level becomes approximately the same as the received signal level due to the presence of the received signal. The above problem occurs. Next, a conventional technique for solving this problem will be described.

FIG. 5 shows a conventional example of an echo canceller employing a sine algorithm. This ζ is the fifth
It is assumed that the circuits in the figure are connected oppositely via a two-wire transmission line 4. If the target cable is a subscriber cable, one is installed on the central office side and the other on the subscriber side. To simplify the explanation, baseband transmission will be assumed here, and FIG. 5 will be explained as a subscriber side circuit.

In FIG. 5, a binary data series is supplied to an input terminal 1, and a transmitter 3 and an adaptive digital filter 8
is input. In the transmitter 3, the binary data series is converted into a transmission line code and then converted into a no hybrid transformer (HY
B) 3 to the two-wire transmission line 4. On the other hand, a part of the transmission signal generated by the transmitter 2 appears as an echo component at the output of the hybrid transformer 3 and is supplied to a low-pass filter (LPF) 5. In addition, the received signal sent from the opposite side (in the current explanation, the station side) facing the circuit in FIG.
The signal is supplied to a low-pass filter 5 via a transformer 3. Therefore, the output of the low-pass filter 5 is a mixed signal containing a received signal and an echo. Note that the role of the low-pass filter 5 is to suppress frequency components outside the desired signal band. The output of the low-pass filter 5 is fed to a subtracter 10. Here, an adaptive digital filter 8, a D/A converter (DAC) 9, a subtracter 10. A closed loop circuit consisting of adder 11, polarity determining circuit 12 and multiplier 13 operates to remove echoes in the mixed signal that is the output of Ronox filter 5.

This is achieved by the adaptive digital filter 8 generating echo replicas. Therefore, the adaptive digital filter 8 will be explained in detail.

FIG. 6 shows a detailed block diagram of the adaptive digital filter 8 shown in FIG.

Input signals 105 and 106 in FIG. 6 are the binary data series (+1
or takes a value of -1) and the output of the multiplier 13. Further, the output signal 107 in FIG. 6 corresponds to the output signal of the adaptive digital filter 8 in FIG. The binary data series 105 is transmitted through the delay element 100
．． , multiplier 101°, 1011 , ..., 1
01R-, and coefficient generator A. ,A8,...
AR. Delay element 1 giving a delay of T seconds
001.100t, 100N/R-1
are connected in this order, and each can be realized by a clip-flop. Here, N and R are positive integers, and R is a divisor of N. Further, the data rate of the binary data series 105 is 1/T bit/sec. Delay element 10
0. (1=1.21 . . . , N/R−t) are respectively output from multipliers 101 . ．． 101. Ya1゜...,1
01. +R-1 and coefficient generator Aj, A, +1. ...
, people, +t, are supplied. However, j=lXR.

To the multiplier 101, l0IK+□, ..., +N to 101
-R (to = 0.1....., R-1) Then, the coefficient generators AK, AK+, ......, AK+
After each coefficient that is the output of NR is multiplied by the input data, all the multiplication results are input to the adder 102K and added. 8 adders 102°, 102+,...
. . , the output of 102R-□ becomes the input contact of the switch 103. The switch 103 is a multi-contact switch with a cycle of T seconds), and the adders 102°, 1021, .
..., 102R-1 are selected and outputted in this order every T/R seconds, resulting in an output signal 107. Output signal 107 is an echo replica! An echo replica is generated every +, T/R seconds.几 is the interpolation constant (interpolation
B is usually an integer of 2 or more in order to remove echo within a required signal band. On the other hand, the switch 104 that operates in synchronization with the switch 103 has its input and output reversed. That is, switch 104
serves to sequentially distribute the input signal 106 to the eight contacts every T/R seconds. Each contact output of the switch 104 is
It is supplied to a coefficient generator present in a signal path that is input to a contact corresponding to a switch 105 that operates synchronously. Next, the coefficient generation circuit will be explained in detail.

FIG. 7 shows the coefficient generation circuit 1 (""Or 1 +
. . , N-1).

The input signal 200 in FIG. 7 is the binary data series 105 or the delay elements 1001, 1oo, . . . in FIG.
・It corresponds to an output signal of 100N/R-s. Furthermore, the input signal 201 in FIG. 7 corresponds to the contact output of the switch 104 in FIG. Furthermore, output signal 203 in FIG. 7 corresponds to the output of coefficient generator A1 in FIG. In FIG. 7, input signals 200 and 2
01 is supplied to the multiplier 204, and the multiplication result is sent to the adder 2.
This is one input of 05.

The output of the adder 205 is fed back through a delay element 206 of T seconds, and the coefficients are updated every T seconds using the input signals 200 and 201 supplied to the multiplier 204.
This is achieved by adding the correlation value of 1 to the coefficient value of one sample before. Output signal 203 is the coefficient.

The echo replica generated by the adaptive digital filter 8 shown in FIG. 5, which has been explained above with reference to FIGS. This becomes one input of the device 10. In the subtracter 10, a low pass filter 5
The difference signal obtained by subtracting the echo replica from the mixed signal (= [echo] + [received signal]) that is the output signal of (= [residual echo] + [received signal]. However, [residual echo] = [echo] - [echo] A replica]) is obtained and supplied to the receiving section 6, adder 11, and amplitude control circuit 14. In the receiving section 6, clock extraction, demodulation of the received signal, etc. are performed, and the identified data appears at the output terminal 7. The amplitude control circuit 14 functions to control the maximum amplitude value of the random signal generated by the random signal generator 15 with reference to the amplitude or power of the difference signal that is the output of the subtracter 10.

The random signal having the maximum amplitude controlled by the amplitude control circuit 14 becomes one input of the adder 11.

The difference signal that is the output of the subtracter 10 and the amplitude-controlled random signal that is the output of the amplitude control circuit 14 are sent to the adder 11.
After the signals are added, the polarity detector 12 detects only the polarity thereof. Furthermore, the output of the polarity detector 12 is transmitted to a multiplier 13
After being multiplied by 2α (α is a positive number), the signal is supplied to the adaptive digital filter 8 as an error signal. Input signal 106 in FIG. 6 corresponds to the error signal. In order for the aforementioned adaptive digital filter 8 to perform an adaptive operation, it is necessary for the polarity detector 12 to correctly detect the polarity of the residual echo.

However, since the received signal is included in the difference signal that is the output of the subtracter 10, in FIG.
Assuming that the output of is directly input to the polarity detector 12,
When the residual echo level reaches the same level as the received signal level,
The polarity of the residual echo cannot be accurately obtained from the output of the polarity detector 12. Therefore, the adaptive capability of the adaptive digital filter 8 will be lost. Therefore, conventionally, as shown in FIG. 5, an adder 11, an amplitude control circuit 14, and a random signal generator 15 are added.
A method has been used in which the adaptive operation of the adaption digital filter 8 is guaranteed by adding a random signal having a level equivalent to the received signal level to the difference signal that is the output signal of the subtracter 10.

This method generates a probability of canceling the received signal (by adding a random signal of the same level as the received signal to the difference signal). Since this probability is the probability that the polarity of the residual echo is correctly obtained by the polarity detector 12, the adaptive operation of the adaptive digital filter 8 is guaranteed.

However, in the conventional method shown in Fig. 5, it is necessary to generate a random signal, and in order to obtain the desired degree of echo suppression, the maximum value of the random signal to be added to the difference signal must be equal to the received signal level. This has the drawback that it requires complicated control to maintain a certain level, and the hardware scale becomes large. Furthermore, since the tap coefficients are updated using the polarity of the error signal, the conventional method employing the sine algorithm has the disadvantage of a long convergence time.

(Object of the Invention) Therefore, an object of the present invention is to provide an echo removal method that is easy to control and requires small hardware.

Another object of the present invention is to provide an echo removal method with short convergence time.

(Structure of the Invention) According to the present invention, an echo replica generated by an adaptive filter on the 4-wire side of a 2-wire/4-wire conversion circuit is used to remove echoes leaking from the transmitting circuit to the receiving circuit. The removal method includes subtracting the echo replica from a mixed signal in which the echo and the received signal are mixed to obtain a difference signal, and then calculating the sum or difference between the difference signal and a delayed signal obtained by delaying the difference signal. to obtain a first error signal, and calculate the correlation between the polarity of the echo replica and the polarity of the difference signal;
A second error signal is generated by adding the polarity of the first error signal to a signal obtained by multiplying the correlation signal by a constant, and the second error signal is fed back to the adaptive filter. An echo removal method characterized by the following is obtained.

(Principle of the Invention) The first point of the present invention is to focus on the characteristics of the receiving eye pattern, and to configure the following so that the probability of receiving signals being canceled does not become zero. In other words, according to the characteristics of the receiving eye pattern of the transmission line code including the binary coding system, the current value and the value l-T seconds (! is a positive integer) ago are almost the same value or have opposite polarity and each absolute The minimum probability that the values are almost the same takes a certain positive value that is not zero. Therefore, for the difference signal (= [residual echo] + [received signal]), the current value and l・T
By taking the difference or sum of the values from seconds ago, the received signal component will be canceled with a probability of a certain positive value that is not zero. Therefore, by detecting the polarity of the difference or sum, the sign of the residual echo can be detected with a probability of a certain positive value other than zero, so that the adaptive operation of the adaptive filter is guaranteed.

The second point of the present invention is that the step size is adaptively changed when updating the tap coefficients of the adaptive filter. The present invention focuses on the fact that when the residual echo is large, there is a strong correlation between the polarity of the pseudo echo and the polarity of the residual echo, whereas when the residual echo is small, there is no correlation between the two. Adaptively change the step size depending on the correlation value. Therefore,
It becomes possible to significantly shorten the convergence time compared to the conventional method.

(Example) Next, the present invention will be explained in detail with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

In this figure, functional blocks given the same reference numbers as in FIG. 5 have the same functions as in FIG. 5. The difference between FIG. 1 and FIG. 5 is that the circuit includes a subtracter 16 and a delay element 17 that provides a delay of T seconds, and
The other configuration is exactly the same as that of FIG. 5, except for the presence/absence of , and the circuit consisting of polarity detectors 19 and 23, correlator 20, and multiplier 21. Before explaining these differences, the overall configuration will be briefly described.

The binary data series supplied to the input terminal 1 is supplied to the transmitter 2 and the adaptive digital filter 8. After the binary data sequence is converted into a transmission line code in the transmitter 2, it is sent to the two-wire transmission line 4 via the hybrid transformer 3. Here, due to the impedance mismatch of the hybrid transformer 3, the output of the transmitter 2 leaks into the receiver circuit as an echo and is supplied to the low-pass filter 5.

On the other hand, the received signal is also supplied to the low-pass filter 5 via the transmission line 4 and the hybrid transformer 3. The mixed signal (=[echo]+[received signal]) whose unnecessary high frequency components have been suppressed by the low-pass filter 5 is supplied to the subtracter 10. Therefore, the pseudo echo (echo replica) generated by the adaptive digital filter 8 is
After being converted into an analog signal by the /A converter 9,
The signal is input to the subtracter 10 via the interpolation filter 22.

Therefore, the difference signal that is the output of the subtractor 10 (= [mixed signal) - (echo reflex'lJ force) = [: ﾗﾞﾞﾞﾞﾞ:
] + [received signal] - [echo replica]) component, residual echo (= (echo) - (echo replica))
If the received signal becomes sufficiently smaller than the received signal, the received signal is accurately demodulated in the receiving section 6, and the received 2
A value data series appears. Note that the interpolation filter 22 is
It functions to suppress harmonic components contained in the output of the D/A converter 9. Here, an adaptive digital filter 8, a D/A converter 9, an interpolation filter 22, a subtracter 10 . Subtractor 16, polarity detector 1
A closed loop circuit consisting of the filter 2 and the multiplier 13 realizes the adaptive operation of the adaptive digital filter 8. The configuration of the adaptive digital filter 8 may be the same as that shown in FIGS. 6 and 7, similar to that described in the conventional example shown in FIG. The output of the polarity detector 12 is multiplied by the output of the multiplier 21 in a multiplier 13 and supplied to the adaptive digital filter 8 as an error signal. Next, the relationship between the output of the polarity detector 12 and the polarity of the residual echo component in the difference signal that is the output of the subtracter 10 will be explained in detail, but before that, the transmission line code will be described.

Figure 2 shows a typical example of a 2-short rod.
, ) show the pi-phase code, and (b) shows the pulse waveform of the M 8 K (minimum shift keying) code, respectively. As shown in FIG. 2(a), in the pi-phase code, pulse waveforms with inverted polarities are assigned to data 0'' and +1111. Both pulses are 1
The polarity is reversed at the center of the pit width T seconds, and the polarity is balanced within one pit. On the other hand, as shown in FIG. 2(b), four types of pulse waveforms are prepared for M8. i.e. “On” and “1”
Two types of pulse waveforms, ■ mode and ○ mode, with reversed polarity are prepared for the data. These two types of mode transitions are indicated by thick arrows in Figure 2(b).
The current mode is determined by the mode one bit before. This MSK code has the characteristic that the polarity always inverts at the bit boundary.

Note that in the MSK code, the positive and negative values are balanced within one bit for "1", but the positive and negative values are not balanced for 10''. However, the mode transition in Figure 2 (b) As is clear from the direction of the thick arrow shown, if there is an even number of Os in a continuous bit sequence, the positive and negative values are balanced and the DC component can be almost ignored.The transmission line code shown in Figure 2 is the transmitter 2 in FIG.
It will be output in .

are the biphase code and MS, respectively, corresponding to FIG.
This is a receiving eye pattern of K code. As shown in the figure,
The receiving eye pattern has high frequency components cut off and becomes rounded. Now, pay attention to FIG. 3(a). For each combination of four sample points separated by T seconds (tot(1'
) j (t+j 1%) 9 (t2・t2'
) and (ts·t,′). At this time, t=tm
' If the value obtained by subtracting the sample value of 1=1 m from the sample value of (m=op 1 j2 s 3) is defined as Affl, it can be seen that Am is given as shown in Table 1.

Table 1. Values of Am “0” and “1” in case of Bi7Aze code
If we assume that the probability of appearance of “ is 1/2, then A0=
Table 1 shows the probability that A, =O, A, =O and A, =0.
The values are 1/4, 1/4, 1/2, and 1, respectively. In this example, we considered four sets of sample points separated by T seconds as shown in Figure 3(a), but as is clear from the figure, no matter what phase they take, apart from positive/negative reversals, Table 1 shows that It can be seen that there are no other patterns than those shown. Therefore, the minimum probability that the value obtained by subtracting the sample value T seconds ago from the current sample value is zero is 1/4. Next, Figure 3(b)
Considering the received eye pattern of the MSK code, Am is given as shown in Table 2 with reference to the mode transition in FIG. 2(b).

Assuming that the probability of occurrence of 0'' and 11'' is equal and 1/2, Ao=0-A, =OtA2=O and A, =
The probability of becoming 0 is 1j1/2.1/4 as shown in Table 2.
and 1/4. In this example, we considered four sets of sample points separated by T seconds as shown in Figure 3(b), but as is clear from the figure, regardless of the j phase, apart from positive/negative reversal, , it can be seen that patterns other than those shown in Table 1 are not acceptable. Therefore, also in the case of the MSK code, the minimum probability that the value obtained by subtracting the sample value T seconds ago from the current sample value is zero is 1/4. As mentioned above using the biphase code and MSK code as examples, the minimum probability that the value obtained by subtracting the sample value T seconds ago from the current sample value is zero is both 1/4. Recognize. When considering transmission line codes other than these codes in the same way, it is clear that the minimum value of the probability has a value other than zero.

Furthermore, up until now, we have been targeting the value obtained by subtracting the sample value T seconds ago (the dental rate is 1/T pits/second) from the current sample value, but now we are using the current sample value! - It can be seen that the minimum probability that the value obtained by subtracting the sample value T seconds ago (! is a positive integer) is zero is also 1/4. Next, the meaning of this probability in the adaptive operation of the echo canceller will be explained with reference to FIG.

In one embodiment of the invention shown in FIG.
is a delay element giving a delay of T seconds, reference numeral 16 is a subtractor, and reference numeral 12 is a polarity detector. Here, in order for the adaptive digital filter 8 to perform an adaptive operation, the difference signal (=[echo]+[received signal]−[echo replica]) which is the output of the subtracter 10 is sent to the polarity detector 12.
As mentioned earlier, it is necessary to have a non-zero probability that the polarity of the residual echo (=(echo > (echo replica)) component included in the residual echo (=(echo > (echo replica)) component is not zero. The element 17 is added to satisfy this condition, and the output of the subtracter 17 is such that the value obtained by subtracting the value T seconds ago from the current value appears. 2, it is clear that the received signal component in the difference signal that is the output of the subtracter 10 becomes zero with a probability of 1/4 or more at the output of the subtracter 16. On the other hand, Considering the residual echo component included in the output of the subtractor 16, the value obtained by subtracting the residual echo value T seconds ago from the current residual echo value is output from the subtractor 16 as the residual echo component. Since there is no correlation between the residual echo value and the residual echo value T seconds ago, T
The value of the residual echo seconds ago can be considered as random noise.

The amplitude distribution of the residual echo values T seconds ago is symmetrical between positive and negative,
The probability that the amplitude d satisfies ldl≦δ (however, 0≦S) is not zero but takes a positive value. Therefore, it can be seen that the probability that the polarity of the current residual echo is accurately outputted by the polarity detector 12 which receives the output symbol of the subtractor 16 as input is a certain positive value that is not zero.

Therefore, the adaptive operation of the adaptive digital filter 8 is guaranteed. In addition, in FIG. 1, the delay element 17 was explained as providing a delay of T seconds, but Table 1
And as mentioned in the explanation of Table 2, the amount of delay is j−
A similar effect can be obtained by setting the time to T seconds (l is a positive integer).

Next, the operation of the correlator 20 shown in FIG. 1 will be explained. The correlation value between the output of the polarity detector 23 and the output of the polarity detector 19 is calculated by the correlator 20 and then multiplied by 2α (α is a constant) by the multiplier 21 and supplied to the multiplier 13 . here,
The polarity of the difference signal (=[residual echo]+[received signal]) which is the output of the subtractor 10 appears in the output of the polarity detector 23, and the polarity of the echo replica appears in the output of the polarity detector 19. Therefore, if we pay attention to the fact that when the residual echo is large, the polarity of the difference signal and the polarity of the echo replica are correlated, but when the residual echo is small, the two have no correlation. 20 outputs a large value when the residual echo is large, and outputs a small value when the residual echo is small.

Therefore, the output of the correlator 20 is scaled by a factor of 2α in the multiplier 21 and used as a step size, and the polarity of the output of the polarity detector 12 is added to this step size, which is then fed back to the adaptive digital filter 8. By doing so, it becomes possible to significantly shorten the convergence time.

FIG. 4 is a block diagram showing another embodiment of the invention. In this figure, functional blocks given the same reference numbers as in FIG. 1 have the same functions as in FIG. 1. The difference between FIG. 4 and FIG. 1 is that the first subtracter 16 is replaced by an adder 18 in FIG. 4, and the other parts are exactly the same. Therefore, in FIG.
Regarding the difference signal that is the output of 0, the sum of the current difference signal value and the difference signal value T seconds ago appears at the output of the adder 18, and the polarity detector 12 detects the polarity of this sum value. It turns out. Therefore, using FIG. 2 showing an example of a transmission line code and FIG. 3 showing an example of the receiving eye pattern, Table 2 and Table 3
Let's find the table corresponding to . First, pay attention to Figure 3 (,), and compare the combinations of four sample points separated by T seconds (
to・10'), (ti j t,' ), (t2 y
tt') and (t3jt3'). At this time, the sample value of "=tm' (m=011t2j3) and t
If the sum of the sample values of ""tm is Bm, then Bm is as shown in Table 3.
It can be seen that it is given as follows. Similarly, Fig. 3(b)
, Table 4 is obtained.

Table 3. Bm value table 4 for pi-phase code. Assuming that the appearance probabilities of Bm values "O" and @1" when M8 is a sign are equal and each 1/2, Bo=0?B, =OjB2=0 and B,
The probability that =O is 1/2, 1/4, 1/2, and 1 in the case of the biphase code shown in Table 3, respectively.
In the case of the MSK codes shown in Table 4, they are 1.1/2 x 1/4 and 1/2, respectively. Therefore, the minimum probability that the sum of the current sample value and the sample value T seconds ago becomes zero is 1/4, and this holds true for any sampling phase. In addition, Tables 3 and 4 show the case of the P7 code and the M8 code, respectively, but if you consider other transmission line codes in the same way, the current sample value and T
It is clear that the minimum probability that the sum with the sample value seconds ago is zero has a non-zero value. Furthermore, it goes without saying that the minimum probability that the sum of the current sample value and the sample value l-T seconds ago (l is a positive integer) is zero also has a non-zero value.

Returning to the explanation of FIG. 4, the difference signal that is the output of the subtracter 10 is supplied to the receiver 6, as well as to the adder 18 and a delay element 17 that provides a delay of T seconds. Further, the output of the delay element 17 serves as one input of the adder 18. Therefore, the sum of the current value and the value T seconds ago appears at the output of the adder 18. From Table 3 and Table 3, it is clear that the received signal component in the difference signal that is the output of the subtracter 10 becomes zero at the output of the adder 16 with a probability of 1/4 or more. On the other hand, considering the residual echo component included in the output of the adder 18, the sum of the current residual echo value and the residual echo from T seconds ago is output from the adder 18 as the residual echo component. Since the current residual echo value and the residual echo value T seconds ago are uncorrelated, the residual echo value T seconds ago can be regarded as random noise. The amplitude distribution of the residual echo values T seconds ago is symmetrical in positive and negative directions, and the probability that the amplitude d will be lal≦δ (however, O≦δ) is not zero but takes a certain positive value. Therefore adder 1
It can be seen that the probability that the polarity of the current residual echo is accurately outputted by the polarity detector 12 which receives the output signal of 8 takes a certain positive value that is not zero. Therefore, adaptive
Adaptive operation of the digital filter 8 is guaranteed. In addition, in FIG. 4, the delay element 17 was explained as providing a delay of T seconds, but as stated in the explanation of Tables 3 and 4, the delay amount is expressed as l-T seconds (i is a positive integer). A similar effect can be obtained. Note that the operation of the correlator 20 is exactly the same as that shown in FIG. 1, and it is clear that the convergence time can be significantly shortened.

Although the detailed explanation has been given above based on the embodiment, the line stopper that compensates for the line loss of the two-wire transmission line is considered to be included in the receiving section 6 in FIGS. 1 and 4. Alternatively, it may be inserted between the low-pass filter 5 and the subtracter 100. Furthermore, when the MSK code is adopted, the configuration of the adaption digital filter 8 is different from that in the case of the pi-phase code due to two reasons: the pulse waveforms for "O" and "1" are different, and each has ■ mode and ○ mode. Slightly different. In other words, it is necessary to prepare two types of tap coefficients corresponding to the different pulse waveforms of "0" and "1" and update them individually.In addition, when the transmitter 2 receives the mode signal, it is necessary to prepare two types of tap coefficients and update them separately. It is necessary to make a distinction. Furthermore, in the explanation so far, it has been assumed that the delay amount of the delay element 17 is T seconds or l-T seconds (l is a positive integer), but in practice, it is sufficient that it is around j·T seconds. Needless to say. Further, the interpolation filter 22 is not necessary if the purpose is to remove echo only at sampling points where echo replicas are generated.

(Effects of the Invention) As described in detail above, according to the present invention, the difference signal (=
[Residual echo] + [Received signal]), the current value and! - T seconds (where l is a positive integer and 1/T is the data rate) The received signal component is canceled with a probability of a certain positive value that is not zero (by taking the difference or sum with the previous value). Therefore, by detecting the polarity of the difference or sum, the adaptive operation of the adaptive digital filter is guaranteed.

Further, according to the first and second inventions, A! - By combining a delay element that provides a delay of 'l' seconds and a subtracter or an adder, the above-mentioned adaptive operation can be guaranteed, making it possible to perform echo cancellation without requiring complicated control and with a small hardware scale. I can provide a method.

Further, according to the present invention, since the step size can be adaptively changed depending on the size of the residual echo, it is possible to significantly shorten the convergence time.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the first invention;
Figures (a) and 9 (b) are examples of pulse waveforms of transmission line codes,
Figure 3 (a)? (b) shows an example of a receiving eye pattern, FIG. 4 is a block diagram showing an embodiment of the second invention, FIG. 5 is a block diagram showing a conventional example, and FIG. 6 shows the configuration of an adaptive digital filter. 7 are diagrams showing the configuration of the coefficient generator. In the figure, 2 is a transmitter, 3 is a hybrid transformer, 4 is a two-wire transmission line, 5 is a low-pass filter, 6 is a receiver, 8 is an adaptive digital filter, 9 is a D/man converter, 10 and 16 are subtracters, 11 and 18 are adders, 12
19 and 23 are polarity detectors, 13 and 21 are multipliers,
14 is an amplitude control circuit, 15 is a random signal generator, 17
is a delay element, 20 is a correlator, 22 is an interpolation filter, 10
0, Tsu1002. ...-' 100N/R-1 is a delay element, 101°, 102. TS・・・Problem 101
N-1 is a multiplier, 102o. 102, ,...j102R-, is an adder;
103 and 104 are multi-contact switches, AO+Al1/
. . . AN-1 is a coefficient generator, 204 is a multiplier, reference numeral 205 is an adder, and reference numeral 206 is a delay element. Figure 2・0・(a) ・1-(b) Figure 3(a) (b)

Claims (1)

[Claims] (1) Adaptive on the 4-wire side of the 2-wire/4-wire conversion circuit.
An echo removal method for removing an echo leaking from a transmitting circuit to a receiving circuit using an echo replica generated by a filter, the method comprising subtracting the echo replica from a mixed signal in which the echo and the received signal are mixed. After obtaining the signal, a first error signal is obtained by calculating the sum or difference between the difference signal and a delayed signal obtained by delaying the difference signal, and the correlation between the polarity of the echo replica and the polarity of the difference signal is determined. a signal obtained by multiplying the correlation signal by a constant, giving the polarity of the first error signal to generate a second error signal, and feeding the second error signal back to the adaptive filter. An echo removal method characterized by: