JPS61140231A - Automatic circuit equalizing device - Google Patents

Abstract

PURPOSE:To quicken convergence of circuit characteristic equalizing and heighten its reliability by adjusting time and quantity of renewing correcting coefficient of tap coefficient. CONSTITUTION:An equalizing filter part 100 is connected to the receiving line 12 of a transmission line 10 and equalizing signals YK are inputted to a discriminating and deciding part 14 for code deciding, and deciding codes ak are outputted to output 16. Difference of the equalizing signal YK anddeciding code ak, ek=Yk-ak, is outputted to output 18 of the deciding part 14 and inputted to an estimation controlling part 200. The estimation controlling part 200 estimates fluctuation of characteristic of the transmission line 10 from the error signal ek and corrects tap coefficient of the equalizing filter part 100, and makes automatic equalizing optimum condition. So-called ADA process and FIA process basing on maximum inclination process are applicable favorably to this algorithm of automatic equalizing, that is, tap coefficient adjusting process.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an automatic line equalization device, and more particularly to a repetitive automatic line equalization device for data transmission including image data, such as in facsimile.

As is well known, for example, a transversal filter or acyclic filter used in an automatic equalization layer basically consists of a tapped delay line into which a received signal from a transmission path is input, and a tapped delay line for each It consists of a multiplier that multiplies the output of the tap by a tap coefficient, and an adder that totals the output of this multiplier and outputs an equalized signal.

In such a transversal filter, fluctuations in transmission path characteristics are estimated and tap coefficients are corrected to optimize automatic equalization. This automatic equalization algorithm, or tap coefficient adjustment method, is! &The so-called ADA method and FIA method based on the large slope method are relatively commonly used.

As is known, in the ADA method, the correction value of the tap coefficient is proportional to the partial differential value of the quality evaluation function with respect to the tap coefficient.

Further, in the FIA method, the amount of modification of the tap coefficient is limited to a fixed value, and the sign of the positive amount of tl depends on the partial differential value of the quality evaluation function with respect to the tap coefficient. The H〇^ method is superior in convergence of tap coefficients, but the FlA method is simpler in terms of circuit configuration.

Since the convergence speed of the equalization algorithm depends on the number of averaging operations in the evaluation function and the modification coefficient included in the tap coefficient modification amount, there are systems in which the modification coefficient is not fixed. For example, when starting communication, it is common to train a modem device connected to one line. In that case, when the amount of error is large, such as at the beginning of training, the correction coefficient is set large, and as the correction coefficient converges over an hour, the correction coefficient is reduced, and so on. There is a method of assigning temporal weight to the differential value.

In conventional methods of this kind, the timing and amount of change of the correction coefficient are fixed to a predetermined procedure. Therefore, the correction coefficient becomes small before sufficient equalization is performed, and normal data transmission may not be performed. In addition, normal data transmission systems perform error checks such as parity and CRG. Particularly in facsimile transmission, a system that checks for errors on a line-by-line basis is common. However, the error determination result from such error checking does not necessarily match the error determination result at the equalization layer, and therefore there may be a situation where an error is detected in the system even though the 1 equalization error is sufficiently small. It can occur.

Wataru-ichito The present invention overcomes these drawbacks of the prior art.

It is an object of the present invention to provide an automatic line equalization device that improves convergence in automatic line equalization and increases its reliability.

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention includes an equalizing filter means that includes an acyclic filter and equalizes a received signal from a line and outputs it as an equalized signal; It has a correction control means for correcting the equalization signal, and an identification judgment means for discriminating the equalization signal and outputting it to the host machine. The present invention is characterized by an automatic line equalization device that updates the amount of tap coefficient correction over time in accordance with at least one of the error determination results in the host machine. A detailed description will be given below based on one embodiment of the present invention.

First, referring to the tJSZ diagram, the basic configuration of the automatic equalization layer is shown. Become. This basically consists of a tapped delay line +02 to which 2N+1 (N is a natural number) received signals ak (where -N-N) sampled at time intervals T are sequentially inputted from the transmission line IO, A multiplier 104 that multiplies the output Xj (j-N-N) of each tap by a tap coefficient Cj, and 5+! The adder 110 sums the outputs 106 of the adders 104 and outputs an equalized signal Yk to an output 108.

The equalized signal Yk is input to the discrimination determination section 14, where a sign determination is performed, and a determination code ak is outputted to the output 1B. This determination output is supplied to a host machine, such as a facsimile machine, to which the eye movement equalization layer is connected. Determination unit 1 of D
The difference ek=Yk-5k between this equalized signal Yk and the judgment code 5 is outputted to the output 1B of 4.

It is input to the estimation control section 200.

The estimation control section 200 estimates fluctuations in the characteristics of the transmission path IO from the error signal ek, corrects the tap coefficients of the equalization filter section 100, and optimizes automatic equalization. As the automatic equalization algorithm, that is, the tap coefficient adjustment method, the so-called AnA method and FlA method based on the maximum slope method are advantageously applied.

In the AOA method, when the evaluation function of one quality is D, the y-th modification algorithm of the tap coefficient Cj is based on the following formula.

(ν÷l) (F) (ν)Gj
=C” -aaD/acJ
(1) Also, in the FrA method, the following results.

As will be seen from now on, the amount of correction of the tap coefficient, that is, the second term in equation (1) or (2), is proportional to the partial differential value aO/aCj in the ADA method. Further, in the FIA method, it is limited to i1Δ, and its sign depends on the partial differential value 8cj1acj.

For example, to explain the ADA method, the evaluation function is
It depends on the equalization algorithm, but 9! Taking the minimum error form of the evaluation function Da for several m as an example, it is as follows.

Here, especially when m is 2, the NSE method is used.
2/ a Cjcc Z"X -ek
(4) k-1k-.

However, here It uses the following relationship.

Similarly, when m is 1, the MAE method is used.

The estimation control unit 200 of this embodiment corrects the tap coefficient ICj l using the equalization algorithm of the AIIA method described above. For example, in the ADA MSE method, the correction width of the tap coefficient is ΔGj=α(ΣX −ekl (7
) k k-.

Also, in the 80A MAE method.

ΔCj=ct (Σ
(8) k k-.

becomes.

The modification coefficient α can normally be updated advantageously by time management using a counter timer. In the embodiment of the present invention, adaptive updating is performed by combining this with a convergence determination based on an error exponentiation, for example, the square. This convergence judgment is based on the equalization error signal ek! J! It is carried out by the detective M.

Specifically, Σsk2 is calculated, and the correction coefficient α is updated by referring to the contents of the time management counter mentioned above.

The embodiment of the present invention further utilizes an error determination signal received from a host machine, such as a facsimile machine, to which the automatic equalization layer is connected.

After determining a code error, if the modem is configured to receive rlJ using, for example, an l-bit signal, the modulator and demodulator accumulates this signal at each sampling time and refers to this integral value and the contents of the counter. while updating the correction coefficient α. Note that the updating using the square integral value of the equalization error and the updating using the error determination result from the host device may be performed in combination, or only one of them may be performed.

This shows that the situation shown in Figure 3 can be dealt with effectively. In the same figure, numeral 300 is a plot of the time integral value of the equalization film, and numeral 30
0 indicates a plot of the time integral value of judgment error. As we will see, normally, both the integral value of the equalization film and the integral value of the judgment error, as shown in FIG.
Shikashi, CB) and CG become saturated as time passes.
), one of them may diverge,
Furthermore, as shown in (D), both may diverge.

In cases (A) and (0), it is sufficient to monitor either the equalization film or the judgment error, but in cases (B) and (C), it is sufficient to monitor only one. In such a case, a combination of the two methods is effective.

In the embodiment of the present invention shown in FIG. 1, the tap coefficients are corrected by the ADA 115E method by setting -1 to the number of averaging operations in the evaluation function. Namely.

e=Y-i' (1G)
(ν÷1)<ν) Cj = C''-aIIX F e
(11)J In the circuit configuration of FIG. 1, components similar to those shown in FIG. 2 are designated by the same reference numerals. Sign determination unit 142 that constitutes the identification determination unit 14 and determines the geometric ratio signal Yt
The difference between input and output 108 and 18 is obtained by adder +44,
The output 1B is the squarer 202. and the multiplier 204. Each multiplier 204 is supplied to the tapped delay line 102.
Each tap output 120 of
08. and the equalization filter section 10 through the adder 210.
It is input to the multiplier 104 of 0 as a tap coefficient 1cjl.

A correction coefficient α is input to the power n i 208 from the ROM 212 . The output of the squarer 202 is input to the RON 212 through an adder 214 . This also includes
As mentioned above, the line error determination unit 22 of a host machine such as a facsimile machine to which one automatic equalizer is connected
It is configured to receive an error determination signal from the adder 218 through the adder 218. The read operation of ROM 212 is managed by a counter timer 218, which time-controls all operations of the device, including sampling of received signals. In addition, blocks 122, 230, 23 in the same figure
2 and 234 indicate delay circuits having a delay time equal to the arrival time interval T of the sampled received signal.

As can be seen from the figure, Equation 1 (8) is realized by the multiplier +04 and the adder 110, and the equalized signal Y is outputted to the output 108. The difference between this equalized signal and the output of the sign determination section 142 is obtained by the adder 144.

This realizes equation (10). The calculation of the second term of equation (11) is performed by multipliers 204 and 208, and the calculation of the difference between the result and the first term of the equation is performed by adder 21.
This is realized at 0. In other words, in adder 210, j! ! The difference between the tap coefficient of the extended one sampling period (T) and the output of the multiplier 208 is calculated and output as the current tap coefficient.

In the ROM 212, data of the correction coefficient a is stored at a predetermined storage address according to an update procedure. This storage address is accessed by a counter 218 which sequentially increments in accordance with the system clock and manages time of the device, and the correction coefficient α at that storage location is read out. by this.

For example, as shown in FIG. 4, the correction coefficient α is sequentially corrected as time t passes.

The ROM 212 may also store update data of the correction coefficient α, that is, an update schedule, in case the equalization error amount and/or the determination error amount become abnormal. By doing so, more fine-grained equalization control becomes possible.

Although the illustrated embodiment was based on the ADA method, 1
The present invention is not limited to this.

It is clear from the above explanation that the FrA method is also applied in the same way. The convergence of the characteristic equalization is arrested, and its certainty increases.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram showing an embodiment in which the automatic line equalization device according to the present invention is applied to a facsimile machine M'll, and FIG. 3 and 4 are graphs for explaining the present invention in detail. 14, identification determination section 100, equalization filter section +02. ,,Delay line with tap +44. ,,Adder 200, ,Estimation control unit 204.208. Tosanki 210, Adder 212, ROM 218, Counter

Claims (1)

[Scope of Claims] Equalization filter means that includes an acyclic filter and equalizes a received signal from a line and outputs it as an equalized signal; correction control means that modifies tap coefficients of the filter over time; and an identification/judgment means for discriminating the equalized signal and outputting it to the host machine, and the correction control means is configured to calculate the square integral value of the equalization error for the identification/judgment result of the equalized signal and the error in the host machine. An automatic line equalization device characterized in that the amount of correction of the tap coefficient is updated over time in accordance with at least one of the determination results.