JPH0481015A - Digital variable equalizer - Google Patents

Abstract

PURPOSE:To speedily converge an optimum equalizing characteristic with a small processing amount with out fail by continuously deciding the transmission characteristics of a filter calculation executing means while changing one kind of a parameter. CONSTITUTION:A coefficient calculating means 102 is provided to calculated each filter coefficient 103 by executing function transformation with a function corresponding to each filter coefficient with one kind of a parameter 105 as the input. Further, a filter calculation executing means 104 is provided to execute digital filter calculation based on each filter coefficient 103. Therefore, only by changing one kinds of the parameter 105, the filter coefficient 103, namely, the transmission characteristic of the filter calculation executing means 104 is continuously decided, and the required equalizing characteristics is unequivocally decided. Thus, the optimum equalizing characteristics can securely and speedily be converged with the small processing amount.

Description

[Detailed Description of the Invention] [Summary] High-speed data transmission is possible using a metallic bare cable, etc., which is an existing telephone subscriber line, by using a multi-level pulse signal that can take four levels in the amplitude direction, for example. For example, it is installed in a digital subscriber line transmission interface device for simultaneous transmission and reception in both directions, and is used for digital signal processing to equalize signal distortion whose characteristics change depending on the distance, which occurs from the cable that is the transmission path, or to set the gain. Regarding digital variable equalizers, the purpose of this is to perform digital filter calculations while switching filter coefficients, with the aim of making it possible to reliably and quickly converge to the optimal equalization characteristics with a small amount of processing. In a digital variable equalizer that equalizes a signal, each filter coefficient is calculated by inputting one type of parameter and performing a function transformation using a function corresponding to each filter coefficient; and filter calculation execution means for performing digital filter calculation based on each filter coefficient.

Furthermore, the apparatus is configured to include parameter updating means for sequentially optimizing the value of one type of parameter based on the output error and decision symbol output from the decision feedback equalizer.

[Industrial application field]

The present invention uses existing telephone subscriber lines, such as Metalink bare cables, to simultaneously transmit and receive high-speed data in both directions, using a multi-value pulse signal that can have four levels in the amplitude direction. A digital variable equalizer that is installed, for example, in a digital subscriber line transmission interface device, etc., and uses digital signal processing to equalize signal distortion whose characteristics change depending on the distance, which occurs from the cable that is the transmission path, or to set the gain. Regarding.

[Conventional technology]

2. Description of the Related Art Transmission of modulated digital signals using conventional analog telephone lines and the like has become popular. Here, when performing high-speed digital transmission, the loss frequency characteristic of the signal amplitude (hereinafter simply referred to as "loss characteristic") in the metallic cable connecting the subscriber terminal and the telephone exchange, etc. Equalization processing is essential.

Nowadays, with advances in digital signal processing technology and LSI technology, the trend is to use digital signal processing LSIs (DSPs) to process the amplitude equalizer for equalization processing, including the functions of the balancing network and variable attenuator. It is becoming more and more common.

When compensating for the cable loss characteristics described above with a digital signal processing LSI, a digital filter is used, but the following points must be kept in mind.

Since the distance between the subscriber terminal and the station is not constant, the cable length varies from subscriber to subscriber, thereby causing the loss characteristics of the cable to vary as shown in FIG. 7, for example. Therefore, the amplitude equalizer in the digital subscriber line transmission interface device needs to be a variable equalizer whose frequency characteristics can be varied to accommodate various loss characteristics from the subscriber terminal to the station. . A variable equalizer is an equalizer whose frequency characteristics can be changed by changing the parameters of a function that defines the equalizer in accordance with the cable length.

FIG. 8 shows the configuration of a conventional digital subscriber line transmission interface device.

A signal from a subscriber that has passed through an analog transmission/reception cable 801 is separated by a hybrid transformer 802, and then band-limited by a low-pass filter (RLF) 803 on the receiving side to a frequency band of 1/2 or less of the sampling frequency. Thereafter, it is converted into a digital reception signal by an A/D converter 804 consisting of a modulator 805 and a decimation filter 806, and is input to the DSP 81B. This digital received signal is processed by a subtraction section 807 and an amplitude equalization section (EQL) 8.
08 and a decision feedback equalizer (DFE) 810 (these will be described later), the received digital signal 812 is sent out from the DSP 818.

On the other hand, the digital signal input from the transmission digital signal 813 to the DSP 818 is subjected to necessary digital signal processing (2 (I! multi-value conversion, etc.) through the C0D 815, and then output as a digital transmission signal. In this way, the transmission signal output from the DSP81B is
The signal is converted into an analog signal by an A converter 815, and band-limited by a transmitting low-pass filter (SLF) 816 to only components within a band determined by the sampling frequency. and,
The signal is transmitted from a dry eight circuit (DRV) 817 via a hybrid transformer 802 and from an analog transmission/reception cable 801 to the subscriber.

Here, a part of the transmission signal headed for the analog transmission/reception cable 8'01 goes around as an echo component to the reception side via the hybrid transformer 802, is included in the digital reception signal, and is input to the DSP 818, so that the above-mentioned The echo component needs to be canceled out. To this end, an echo canceller (EC) 811 generates the above component as a pseudo echo component from the digital transmission signal, and a subtracter 807 subtracts this from the digital reception signal, thereby canceling the echo component.

In this case, a decision feedback equalization section 810 is connected to the output of the amplitude equalization section 808 on the reception side, and adaptively controls the generation of pseudo echo components in the above-mentioned fine balance circuit processing section 811 based on the digital reception signal. do.

Each function of the DSP 818 described above is realized as a combination of the hardware of the DSP 818 and a microprogram that operates it.

An amplitude equalizer (EQL) 808 within the DSP 818 is the part most relevant to the present invention. Here, loss equalization (correction) of the digital reception signal is performed for loss in frequency characteristics in the analog transmission/reception cable 801. Furthermore, the amplitude equalization section 808 usually also has an AGC (automatic gain control) function. As shown in Figure 7,
The loss characteristics of a cable become steeper as the distance increases, but the loss at low frequencies also increases, so it is necessary to increase not only the slope but also the gain. This is the AGC function.

In the conventional amplitude equalization section 808, three to five types of filter coefficients are prepared according to the cable length of the analog transmission/reception digital 801, and these are used by switching. As a result, the amplitude equalizer 808 is provided with an equalization characteristic (frequency transfer characteristic) as shown in FIG. 9, which corresponds to the loss frequency characteristic shown in FIG. 7, for example, in accordance with the cable length.

At this time, the square sum calculation section 809 in FIG. 8 calculates the square sum of the amplitudes of the signal input to the amplitude equalization section 808, that is, the power, at each predetermined period, and calculates the filter coefficient, that is, the equalization characteristic, from the magnitude. A selection method has been used. For example, a period in which the sum of squares is small indicates that the amplitude of the signal is small overall, so a filter coefficient with a large low-frequency gain and steep high-pass characteristics that corresponds to long cable distances is required. used.

:Problem to be Solved by the Invention] In the above-mentioned conventional example regarding the amplitude equalization unit 808, several types of filter coefficients are selectively used, so accurate correction is performed for small changes in cable length. I can't. In other words, when the required equalization characteristic does not correspond to any of the prepared equalization characteristics, and the characteristic is intermediate between the two, there is a considerable error remaining in the digital reception signal of the equalizer. It has the problem of being stored away.

Also, since it is necessary to calculate the sum of squares, the DS
This has the problem that the processing load in P818 increases.

An even bigger problem is that in the case of digital signals, calculating the amplitude from the sum of squares is not accurate unless the sampling phases match, so it is necessary to repeat the calculation several times until the phases match.

To solve the above problem, a digital variable filter has been proposed that calculates filter coefficients by converting parameters related to cable length using a special conversion function, but because the conversion function is special, it cannot be applied to transversal filters. . Furthermore, the applicant of the present application has filed patent application No. 2-53
No. 787 proposes a digital variable isolator that can be applied to transversal filters; The problem is that it cannot be done.

An object of the present invention is to make it possible to reliably and quickly converge to an optimal equalization characteristic with a small amount of processing.

Means for Solving the Problem] FIG. 1 is a diagram of the present invention. The present invention is based on a digital variable equalizer that equalizes the signal 101 by performing digital filter operations while switching the filter coefficients 103. The equalizer is provided, for example, in a digital subscriber line transmission interface device that performs simultaneous two-way communication using multilevel pulse signals.

First, it has a coefficient calculation means 102 that calculates each filter coefficient 103 by inputting one type of parameter 105 and performing a function conversion using a function corresponding to each filter coefficient. Here, one type of parameter is, for example, a value corresponding to the cable length of the cable through which the signal 101 to be equalized is transmitted. On the other hand, the above-mentioned function is approximately determined, for example, by providing a plurality of sets of one type of parameter value 105 and the corresponding filter coefficient 103 value. Further, these functions are defined as, for example, an n-linear polynomial or an exponential function that uses the value 105 of one type of parameter as an input variable.

Next, it has a filter calculation execution means 104 that executes a digital filter calculation based on each filter coefficient 103. The means is, for example, a transversal filter.

In addition to the above configuration, the present invention can further take the following configuration.

That is, when the decision feedback equalizer (106) is connected after the digital variable equalizer that is the object of the present invention, the value of the one type of parameter 105 described above is changed to the decision feedback equalizer 106. The apparatus further includes a parameter updating means 109 for sequentially optimizing based on the output error 107 and the decision symbol 108 output from the apparatus.

[Function: In the present invention, the filter coefficient 103, that is, the transfer characteristic in the filter calculation execution means 104 can be determined continuously by simply changing one type of parameter 105.
The required equalization characteristics can be uniquely determined, the residual error of the signal after equalization can be reduced, and as a result, the signal error rate can be kept low.

Furthermore, when the decision feedback equalizer 106 is connected to the subsequent stage, the parameter updating means 109 updates one type of parameter 105 to the decision feedback equalizer 106.
Output error 107 and judgment symbol 10 output from 06
By sequentially converging to the optimal value using
The amount of processing is reduced, so it is expected that the hardware will be downsized, and a digital variable isolator that converges reliably and quickly can be realized. In particular, in the present invention, one type of parameter 10
By changing 5, the related filter coefficients 103 are simultaneously updated in association with each other, so that the convergence speed is much faster than in the conventional example 2 in which each filter coefficient is updated independently.

In addition, unlike the conventional example, there is no need to calculate the sum of squares in the process of determining the filter coefficients 103, and the same algorithm as the coefficient correction process in the decision feedback equalizer 106 can be used, simplifying the process. This can simplify digital signal processing.

〔Example〕

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

FIG. 2 is a block diagram of an embodiment of a digital subscriber line transmission interface device according to the present invention.

In the figure, parts given the same numbers as in the conventional example of FIG. 8 have the same functions, so their operations will be omitted.

The difference from the conventional example shown in FIG. 8 is that a digital amplitude variable equalization section 201 is provided in place of the amplitude variable equalization section 808 and square sum calculation section 809 shown in FIG. Further, the output of the decision feedback equalizer 810 (DFE) is not only input manually to the echo canceller (EC) 811 but also input as coefficient correction information 204 to the coefficient converter 203 in the digital amplitude variable equalizer 201. It is a point.

Next, first and second embodiments of the digital amplitude variable equalizer 201 shown in FIG. 2 will be sequentially described.

FIG. 3 is a block diagram of a first embodiment of the digital amplitude variable equalizer 201 shown in FIG.

First, the filter section 202 includes a delay element 301 that delays the filter input by one sample, a delay element 302 that further delays the output of the delay element 301 by l samples, a multiplier 303 that multiplies the filter input by a filter coefficient a0, and a delay element 302 that delays the filter input by one sample. a multiplier 3 that multiplies the output of the element 301 by a filter coefficient a;
04, consists of a multiplier 305 that multiplies the output of the delay element 302 by a filter coefficient a2, and an adder 306 that adds the outputs of the multipliers 303 to 305 and outputs the result as a filter output, and is a 3-tap, second-order transformer. Configure versatile filters. In addition, the coefficient conversion unit 203 includes the coefficient calculation unit 30
7.308.309 and a parameter update unit 310.

The operation of the first embodiment of the digital amplitude variable equalizer 201 described above will be described below.

First, the transfer function of the filter section 202 in FIG.
-1)"ao"al Z-'+a2 Z-2"'
(1) However, z-'=exe(jwT) W=angular frequency T: expressed as baud rate period. Here, the filter coefficients aoba+, al
If expressed as a function of the parameter X, the characteristics of this transversal filter can be continuously changed by changing the value of X.

Here, if the equalization characteristic to be realized in the filter section 202 is, for example, the transfer characteristic shown in FIG. 9 corresponding to the loss frequency characteristic shown in FIG. However, the loss especially at low frequencies does not increase much even if the distance becomes long. Such characteristics can be expressed as one parameter χ (
2) Each function f in Eq. , fl, and f2. Consider implementation using sal and a2.

How to find each function 1°, fl, f2 - Do the following.

First, a considerable number of target equalization characteristic curves are determined using, for example, the cable length as a parameter, a filter is designed with a fixed order, and the filter coefficients are determined. Here, a considerable number means the function f
This means that the number is sufficient to correctly determine 0, fl, and f2. Now, the filter coefficient a of the 3-tamp transfer heral filter in equation (1) corresponds to the equalization characteristics of each of the six cable lengths shown in FIG. sal and a2 are determined using an approximation program for designing a digital equalizer. Each filter coefficient aO+, alt, a2i (1≦1≦
Let 6゛) be .

Next, for example, if the distance is X, ao=fo(x)=a=bx+cx"+dx3=ex'
.

Then, if the length of each cable in the 9th cycle is x, (1≦, ・(3)≦6), calculate the coefficients a, b, c, 'd, e, [ of equation (3)]. In this case, the number of equations required is greater than the coefficients (or
a, b, c on the condition that equation (4) holds true.
, d, and e. The least squares method may be used as a method for finding this.

The coefficients of the functions f5 and f2 in equation (2) can also be determined by the same method as in equations (3) and (4) above. What should be noted here is that if the form of the function in equation (3) is too simple, there will be a large amount of error, and even if a good near-field function is obtained at a certain distance like x1 to x6, at intermediate distances (3) The characteristics of 2 are far from the actual standard curve.
There is a possibility that Conversely, when the number of terms in the function of equation (3) is increased, the actual filter coefficient a is calculated from the parameter X. ,
A problem arises in that the amount of processing required to obtain l az increases. Therefore, the number of terms in equation (3) is selected by cutting and trying several items and selecting the best one. Note that if the number of terms is increased to a certain extent, the characteristics of the function f can be varied continuously. , f3, and f2 are possible.

Corresponding to the equalization characteristics of each of the six cable lengths in Figure 9 (
The results of actually determining the functions f0, fl, and f2 in equation (2) to obtain the filter coefficient 80, als''a2 of the filter in equation (1) are as follows.

For the equalization characteristic shown by the solid line in Fig. 4, which is the same as the standard for the equalization characteristic shown in Fig. 9, the transfer characteristic of the filter of equation (1) using the filter coefficient expressed by equation (5) above is: As shown by the broken line in Figure 4, it can be seen that a transfer characteristic that satisfies the standard can be obtained.

Therefore, in FIG. 3, each coefficient calculating unit 307.3'0 of the coefficient converting unit 203
8 and 309 and performs the calculation of equation (2) using X as a parameter, the filter section 202 shown in the figure can realize the desired equalization process.

Here, the parameter X is the decision feedback equalizer 8 in FIG.
The parameters are generated in the parameter updating unit 310 based on the coefficient modification information 204 from 10, and how to obtain this information will be described later.

Next, FIG. 5 shows the digital amplitude variable equalizer 2 of FIG.
FIG. 2 is a configuration diagram of a second embodiment of No. 01.

In the case of this embodiment, first, the configuration of the filter section 202 is such that the frequency is 80kHz compared to the gain at OHz.
A high-pass filter 501 having a fixed characteristic and having a high-pass characteristic with a gain of about 1 OdB is provided at the front stage. A delay element 502 delays the output by one sample, and a filter coefficient a is applied to the output of the high-pass filter 501. a multiplier 503 that multiplies the output of the delay element 502 by a filter coefficient Fea+;
．． A two-tap, first-order transversal filter consisting of an adder 505 that adds the respective outputs of 505 and outputs the result as a filter output is provided at the subsequent stage. In addition, the coefficient conversion unit 2
03 is composed of coefficient calculating sections 506 and 507 and parameter updating section 508.

In the same manner as in the case of the first embodiment shown in FIG. The results of actually determining each function f0 and fl were as follows.

However, the filter coefficient 80 determined by the above formulas (6) and (7)
The transfer characteristic obtained by combining the transfer characteristic of the filter in the filter section 202 in FIG. This does not strictly match the transfer characteristic of equation (5). When actually viewed in the frequency domain, there is a deviation. However, the processing in the digital amplitude variable equalizer 201 of the digital subscriber line transmission interface device shown in FIG.
, due to the loss in the analog transmitting/receiving cable 801, which is the transmission path, and the high-frequency cut-off characteristics, the shape becomes small and dull and flat ↓ Regenerate the original pulse response waveform with less code amount interference from the resulting isolated pulse waveform response Since the purpose is to do so, they do not necessarily have to be exactly the same in the frequency domain. Here, in the case of equations (6) and (7), when the cable length changes from 0 to 7.5- (corresponding to No. 911), the parameter y must be between -0.45 and 0.95. Since it changes linearly, the coefficient is 80, and as a parameter of al,
It is clear that y may be used instead of . However, since equation (6) includes the calculation of an exponential function, it requires more effort than equation (5) to obtain the filter coefficient from the parameter y. Actual digital signal processing LSI
In the (DSP) processing, as will be described later, it is preferable to divide the value of the parameter y into sections and obtain each coefficient on the assumption that the coefficient changes linearly within each section.

From the above, in FIG. 5, each coefficient calculation unit 506 and 5 of the coefficient conversion unit 203
07 and filter coefficient 80 with y as a parameter.
By determining and al, the filter section 202 in the same figure
It is possible to realize equalization processing that is sufficient for practical use.

In the first embodiment shown in FIG. 3 or the second embodiment shown in FIG.
Parameters X or y, which are input to 309 (FIG. 3) or 506, 507 (FIG. 5), are input to the parameter updating section based on the coefficient correction information 204 from the decision feedback equalization section 810 of FIG. 310 (FIG. 3) or 508 (FIG. 5). The method of this generation will be explained below.

When manually controlling the values of these parameters from the outside, the desired characteristics can be obtained by inputting the X or y value corresponding to the cable length, as shown in (5), (
This is clear from equation 6). However, in a digital subscriber line transmission interface device such as that shown in FIG. 2, it is required that the equalization characteristics be automatically optimized during training sessions.

In this embodiment, the digital amplitude variable equalizer 20 shown in FIG.
The same method as that used for optimizing the tap coefficients of the echo canceller 811 inserted immediately before 1 and the decision feedback equalizer 810 inserted immediately after is used. That is, let xk be the value of Using the polarity 318 (a,) and the error polarity sign (e k), xk-+=xk-a ・sign (ak) Hsign
(em)...Sequentially optimize X or y according to the update formula (8). However, α is a small positive number. Regarding the technology for optimizing flower vases, for example, see IEICE ``Digital Signal Processing,'' p. 224-p. 2.
50.

Now, if the vase settings are at their optimal settings,
The value of sign(a,)·sign(em) in equation (8) is O on average. Here, 0 on average means that each case has a positive or negative value, but over hundreds of cases, the number of times it is positive and the number of times it is negative are almost equal. If the gain of the digital amplitude variable equalizer 201 is insufficient, for example, the number of negative values will be larger on average than the number of positive values, and X or y will become larger. When the optimum value is exceeded, the number of positive and negative changes becomes equal and X or y stays around that value.

The parameter modification algorithm of equation (8) can be strictly determined by equation (5) or (6): or, as shown in FIG. In a range where the characteristics monotonically increase or decrease with respect to changes in , good results can be obtained as a first-order approximation using either equation (8) or equation (2) in which the negative sign in equation (8) is changed to a positive sign.

Figure 6 takes the value of y as an example, and its initial value is 0.37.
This is a computer simulation of how much time it takes for the value of y to converge when the cable length is changed in three ways in the system of the embodiment shown in FIG. As can be seen from the figure, the value of y almost converges after about 600 to 1000 processing times.

In this case, since the decision feedback equalizer 810 is also operated at the same time, its initial value ↓ will also cause a difference in convergence speed. In the example shown in Figure 6, the same initial value is used for three cable lengths. are doing.

Based on the above parameter updating principle, the decision symbol (for example, the four values of +3.+l-1.-3 in a digital subscriber line transmission interface device with North American specifications) that is the output result of the decision feedback equalizer 810 is By inputting the sign and the sign of the error to the parameter updating section 310 (FIG. 3) or 508 (FIG. 5) in the coefficient conversion section 203 as the coefficient correction information 204, the updating section (8) Or the value of X or y is updated sequentially using a similar formula. Then, the updated value of X or y is calculated by the coefficient calculation unit 307~
309 (Figure 3) or 506. By repeating the process of filtering the signal, the value of X or y can be driven to the optimum value.

In addition, in the death world such as judgment return for the calculation of (8) 2, C+, k-, = ci, k - a ・ sig
n(am-+) Hsign(e*)...(9
) However, C41, is the value at time of the j-th dump coefficient of the filter in the decision feedback equalizer I -12... m m: filter order of the decision feedback equalizer ak-+ : time (k- i) Decision symbol α: operations of minute values are performed in a number equal to the filter order m of the decision feedback equalizer. Comparing equations (8) and (9), (8
It can be seen that the equation ) corresponds to the case where i=0.

Therefore, the calculation of equation (8) corresponds to the tap coefficient of the 0th filter.

Due to the similarity between equation (8) and (9), it is easy to incorporate the processing of equation (8) into the decision feedback equalizer, and in that case, the parameter updating means of FIG. Parameter update unit 310 in the figure, parameter update unit 508 in FIG.
etc. will no longer be necessary. An embodiment of a conventional decision feedback equalizer and an embodiment of a decision feedback equalizer incorporating the parameter update means of the present invention are shown in FIGS. 10 and 11, respectively. 1st
Comparing FIG. 1 and FIG. 10, it can be seen that the parameter updating process of the present invention only increases the number of processing terms in the decision feedback equalizer by one.

In the above embodiment, since a process of coefficient correction in the coefficient conversion unit 203 is required, the function of equation (2) is configured as (3) so that the number of terms is as small as possible and the order is as low as possible.
It is necessary to determine a, b, c, ... e of 2. In order to reduce the amount of processing necessary for this coefficient correction process, for example, when the order exceeds a certain level, a conversion table may be provided in a ROM (not shown) or the like.

When using ROM, as described in the case of equation (6), each characteristic of equation (2) is approximated by a line and graphed, and the values on both sides of the line graph for the input value of The coefficients can be obtained by performing two multiplication processes: reading from and performing interpolation processing.

Also, although it will delay the convergence to some extent, the coefficient conversion unit 2
Coefficient calculation units 307 to 309 (Fig. 3) or 50 in 03
It is also possible to divide the coefficient calculation process in 6.507 and modify only 1/n (≧2) coefficients in each processing cycle.

In this case, each coefficient of the filter is determined from a different value of X, but the amount of change in X per cycle is 1/1
Since it is about 000 or less, even if the cycle is off by 1.2 times,
Or, since the value of y only changes by about 2/1000 at most, there is no problem.

Also, the filter coefficient a is calculated as shown in equation (5). After calculating , it is possible to use the result to calculate a and a2, thereby reducing the amount of processing.

J Effects of the Invention] According to the present invention, the transfer characteristic in the filter calculation execution means S: can be determined continuously by simply changing one type of parameter, so that the required equalization characteristic can be uniquely determined. This makes it possible to reduce the residual error of the signal after death, and as a result, it becomes possible to suppress the signal error rate to a low level.

Furthermore, when a decision feedback equalizer is connected to the subsequent stage, the parameter updating means uses one type of parameter as the output error output from the decision feedback equalizer and the decision symbol. By sequentially converging to the optimal value using .

In particular, in the present invention, by changing one type of parameter, related filter coefficients are updated in association at the same time, so the convergence speed is much faster than in the conventional example in which each filter coefficient is updated independently. It also has effects.

In addition, as in the conventional example, there is no need to calculate the two-to-sum in the process of determining filter coefficients, and the same algorithm as the coefficient correction process in a decision feedback equalizer can be used, simplifying the process and making digital It becomes possible to simplify signal processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the present invention, FIG. 2 is a block diagram of an embodiment of a digital subscriber line transmission interface device according to the present invention, and FIG. 3 is a first embodiment of a digital amplitude variable equalizer. FIG. 4 is a diagram showing an example of the equalization characteristics of the digital amplitude variable equalizer according to the present embodiment, and FIG. 5 is a configuration diagram of the second embodiment of the digital amplitude variable equalizer. , Fig. 6 is a diagram showing an example of convergence of the parameter y, Fig. 7 is a diagram showing an example of loss frequency characteristics of a cable, Fig. 8 is a configuration diagram of a conventional example, Fig. 9 is a diagram, etc. FIG. 10 shows an example of a conventional decision feedback equalizer, and FIG. 11 shows a decision feedback equalizer including a parameter calculation section for a variable equalizer. 101... Signal, 102... Coefficient calculation means, 103... Filter coefficient, 104... Filter operation execution means, 105... Parameter, 106... Decision feedback equalizer, 107... - Output error, 108... Judgment symbol, 109... Parameter updating means. Patent applicant Fujitsu Limited

Claims (1)

[Scope of Claims] 1) In a digital variable equalizer that equalizes a signal (101) by performing a digital filter operation while switching filter coefficients (103), each of the filter coefficients (103) is Type parameter (105)
Coefficient calculation means (102) that calculates by inputting and performing function conversion with a function corresponding to each filter coefficient.
A digital variable equalizer comprising: a filter calculation execution means (104) for performing a digital filter calculation based on each of the filter coefficients. 2) A decision feedback equalizer (106) is connected to the latter stage, and performs digital filter calculation while switching the filter coefficients (103), thereby converting each of the filter coefficients (103) of the signal (101) into one type. Parameters (10
5) as input and performs function conversion using a function corresponding to each filter coefficient.
2), and the value of the one type of parameter (105) is calculated from the output error (107) output from the decision feedback equalizer (106).
) and a determination symbol (108), and a filter operation execution means (104) for executing a digital filter operation based on each of the filter coefficients (103). Features a digital variable equalizer. 3) The parameter changing means (109) is included in the decision feedback equalizer (106) so that parameters for calculating filter coefficients are directly obtained as an output of the decision feedback equalizer. Decision feedback equalizer and claim 2
Digital variable equalizer as described. 4) In addition to the processing of a normal decision feedback equalizer, the decision feedback equalizer performs the following:
ign(e_k) or x_k_+_1=x_k−α・a_k・sign(e_
k) or x_k_+_1=x_k-α・a_k・e_k where x_k: Parameter value at time k a_k: Judgment symbol at time k e_k: Output error α: Small real number sign(A): -1 when A<0, A 3. The digital variable equalizer according to claim 2, wherein when ≧0, a process of 1 is performed and the value is output. 5) The function corresponding to each of the filter coefficients is approximately determined by providing a plurality of sets of the value of the one type of parameter and the value of the filter coefficient corresponding thereto. , 2, 3 or 4. 6) Claims 1, 2, 3, and 4, wherein the filter calculation execution means is a transversal type filter.
or the digital variable equalizer described in 5. 7) The function corresponding to each of the filter coefficients is defined as an n-th order linear polynomial whose input variable is the value of the one type of parameter.
6. Digital variable equalizer according to 5 or 6. 8) Claims 1, 2, 3, 4, 5, 6, wherein the function corresponding to each of the filter coefficients is defined as an exponential function whose input variable is the value of the one type of parameter.
or the digital variable equalizer described in 7. 9) Claims 1, 2, 3, 4, 5, 6, and 7, wherein the one type of parameter is a value corresponding to a cable length of a cable through which the signal to be equalized is transmitted.
or the digital variable equalizer described in 8.