JPH0693598B2 - Decision feedback equalizer - Google Patents

Description

The present invention relates to a decision feedback equalizer used for equalizing transmission line waveform distortion in digital transmission.

(Prior Art) FIG. 4 shows a configuration of a conventional decision feedback equalizer. The main part of this decision feedback equalizer is a transversal wave device (FIR).
Wave device or non-recursive wave device), and the judgment output is given to its input. The tap weight addition output of the transversal wave device is a replica of transmission path distortion (intersymbol interference), which is subtracted from the reception input to determine the transmission code by obtaining a waveform without distortion.

(Problems to be solved by the invention) However, in actual digital transmission, due to the influence of the low-frequency cutoff characteristic at the baseband (due to line transformer, amplifier AC coupling, etc.), the impulse response of the transmission line is As shown in Fig. 5, there are cases where "the hem is pulled long". For example, assuming a first-order low-frequency cutoff characteristic, this long-lasting waveform has a negative exponential function. The problem of the decision feedback equalizer using a non-recursive wave device is that the number of taps, that is, the time length of the response waveform that can be generated is finite, so if the received input continues for a long time, This is because it is not possible to compensate the intersymbol interference in the portion before the time corresponding to the tap length. In other words, a large number of taps are required to sufficiently remove the waveform distortion that is long in time.

Therefore, an object of the present invention is a recursive wave device (IIR wave device).
By combining the time t as shown in FIG.
It is an object of the present invention to provide a decision feedback equalizer capable of removing a long “tailing” waveform distortion with> N · T (N corresponds to the number of non-recursive wave taps, T is a symbol interval).

(Means for Solving the Problems) A decision feedback equalizer of the present invention is a transmission line distortion equalizer by decision feedback, which inputs an output value obtained by subtracting an estimated value of the transmission line distortion from a received signal, A code discriminating means for discriminating a code and issuing a discrimination output, and outputting a result obtained by subtracting the discrimination output from the result output as a discrimination error, a non-recursive filter having the discrimination output as an input, The output of the final tap of the cyclic filter is input as the first input of the first adder via the first delay element having a 1-symbol interval,
A first recursive filter for delaying the output of the first adder by one symbol interval by a second delay element, and feeding back the result obtained by multiplying the internal tap weight as a second input of the first adder. And an output of the first adder is multiplied by an external tap weight controlled by a correlation value between the output of the first adder and the determination error, and the multiplication result is delayed by a third delay at one symbol interval. Is input as a first input of a second adder via an element, the output of the second adder is delayed by one symbol interval by a fourth delay element, and the result obtained by multiplying by the internal tap weight is obtained by the first tap. And a second recursive filter that feeds back as the second input of the second adder, and thus the estimated value of the transmission path distortion is the output of the multiplication result obtained by multiplying the external tap weight by the non- Controlled by combining with the output of the recursive filter, Internal tap weights is characterized in that so as to be controlled by the correlation value between the decision error and the output of the second adder.

(Embodiment of the Invention) FIG. 1 shows a configuration of an embodiment of a decision feedback equalizer according to the present invention. Considering the above-mentioned negative exponential response, the order of the recursive wave device is first-order, which is the simplest. Therefore, FIG. 1 is shown as a case corresponding to this example. In this figure, 1, 2, 3, 4 and 5 are shift registers, 11, 12 and 13 are tap weights of an acyclic wave filter, 21 is an external tap weight of a first cyclic filter, and 22 is a first cyclic filter. Internal tap weight of type filter,
31 and 32 are adders, 41 and 42 are subtractors, and 6 is a transmission code judging device. The input 101 of this decision feedback equalizer is a reception input including intersymbol interference, and this is written as y _{n in the} sample value system.
The tap weight of the non-recursive wave device is C _k (k = 1, 2, ..., N-1)
And the tap weights of the first cyclic wave device are K and a, respectively. On the other hand, the intersymbol interference replica 102 is γ _{n in the} sample value system, and the reception signal 103 after the intersymbol interference is removed is
_Let z _{n be} the decision output 104 be d _n, and the decision error 105 be e _n .

By this, Is obtained. However, u _n is the output of the first recursive wave device and follows the following recurrence formula.

u _n = d _nN + a · u _n-1 (2) This recurrence formula corresponds to the first-order IIR wave device 1 / (1-az ^-1 ).

Next, _let z _n = y _n −γ _n (3) e _n = z _n −d _n = y _n −γ _n −a _n (4), and _let the mean square error be ε, Becomes

Now, in order to determine the tap weights C _k , K, a of the non-recursive wave device and the first cyclic wave device, ε is minimized (LM
Called S algorithm), the following partial derivative is obtained.

At Although, recurrence formulas regarding previous u _n from (2), considering that it is a function of u _n-1 is also a, Becomes Now Then, the recurrence formula To get Therefore, Becomes E on the right side of equations (6), (7) and (9)
_{[E n d nk], E} [e n u n] and E [e _n (p _n], the first circuit described above for generating .p _n showing the correlation value of the error signal and the control signal, respectively It has the same configuration as
This second recursive filter comprises a shift register 205, a first
The internal tap weight 222 having the same value a as the internal tap weight 22 of the recursive filter and the adder 232. The input 106 (w _n = K u _n ) is received via the shift register 206, and the output 107
Emits (p _n ).

The receive input y _n can usually be written as:

Here, a _n is a transmission code, g _n is a transmission path impulse response (including waveform distortion), and g _o (11) and g _{N + m} = g _n ρ ^mm m = 0,1, ..., ∞ ... (12) is assumed. Here, ρ is the attenuation constant (0 <ρ <1) of the negative exponential waveform derived from the first-order low-frequency cutoff.

In an actual adaptive decision feedback equalizer, the tap weight is LMS because the above partial derivative is proportional to the above correlation value.
The algorithms 241 and 242 are used to sequentially correct and converge as follows, and adaptive control is performed.

C _k ^{(v + 1)} = C _k ^(v) + α _{Ce n} a _nk (k = 1, 2, ..., N-1) ...
(13) K ^{(v + 1)} = K ^(v) + α _K e _n u _n (14) a ^{(v + 1)} = a ^(v) + α _a e _n p _n (15) where α _C , α _K and α _a are correction coefficients.

FIG. 2 is a diagram illustrating a circuit for obtaining the above equations (13) to (15), that is, circuits 241 and 242 for realizing the LMS algorithm. This circuit is designed by NAM Verhoeckx etc. Digital E
Cho Cancellation for Baseband Data Transmission ", IEEE Transaction on ASSP, vol.ASSP−27, No.6, p
It is a circuit that summarizes several expressions of this kind in p.768-781, December 1979 in a basic form. By taking the cumulative sequentially multiplied by the coefficients using this circuit, it is possible to ^{Z (v + 1) = Z} (v) + αx n e n ... (16) consisting of the general formula calculation. α, x _n , Z ^(v) as α
_{If K} , u _n , and K ⁽ⁿ⁾ are used, it can be applied to the circuit 241, and α, x _n , Z
^If α ^a , p _n , and a ⁽ⁿ⁾ are used as ^(v) , it can be applied to the circuit 242. In this embodiment, v in parentheses is n.

The characteristics of the decision feedback equalizer obtained by the above-mentioned configuration and adaptive control of tap weights are as follows. As an example, consider the case where the waveform of the received input has a negative exponential response corresponding to the low cutoff frequency of 1/200 of the code transmission rate (ρ = e ^{-2π / 200} ), and the main part of the waveform distortion is It can be removed by a non-recursive wave filter with N = 16 taps. Also, the impulse response at t = NT is g _N =
Set to 0.005. The tap weight of the cyclic wave filter capable of removing the intersymbol interference at the "tail" of such a waveform is naturally K = 0.005 a = 0.96907 (= ρ).

Figure 3 shows the results of investigating how the tap weight is correctly controlled and ε decreases (convergence process) by computer simulation. In this simulation, a _n is +1,
Or it is a random sequence of -1, and the * mark in the figure It is an intersymbol interference value. For reference, the mark shows the result when a decision feedback equalizer consisting of only a non-recursive wave device is used. By comparing the two, the effect of the added recursive wave filter is clear.

(Effects of the Invention) As is apparent from the above description, according to the present invention, the first cyclic wave device is connected in series to the non-cyclic wave device,
Furthermore, by combining the output side of the first cyclic wave filter with the second cyclic wave wave filter having the same structure as the wave wave filter, the code can be coded with a simple structure even when the reception input continues for a long time. Interference can be compensated well, and signal quality in digital transmission can be improved, which is a great effect.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a decision feedback equalizer according to the present invention, FIG. 2 is a view showing a basic type circuit of a known LMS algorithm, and FIG. 3 is a simulation result of an embodiment of the present invention. FIG. 4 is a configuration diagram showing an example of a conventional decision feedback equalizer, and FIG. 5 is a graph showing an example of impulse response waveform of a transmission line. In the figure, 1 to 5 are shift registers, 6 is a transmission code determiner, 11 to 13 are tap weights of acyclic wave filters, and 21 and 22 are first
Tap weights of the cyclic wave filter, 31, 32 are adders, 41, 42 are subtractors, 205, 206 are shift registers, 222 is a tap weight of the second cyclic wave filter, 232 is an adder, and 241, 242 are This is the circuit that implements the LMS algorithm.

Claims (1)

[Claims] 1. A transmission line distortion equalizer based on decision feedback, which receives a result output obtained by subtracting an estimated value of the transmission line distortion from a received signal, judges a sign and issues a judgment output, and outputs the result output. To a code discriminating means for outputting a result obtained by subtracting the decision output from the decision output as a decision error, an acyclic filter having the decision output as an input, and an output of the final tap of the acyclic filter at an interval of 1 symbol. Input as the first input of the first adder via the first delay element, the output of the first adder is delayed by one symbol interval by the second delay element, and the result is multiplied by the internal tap weight. To a second input of the first adder, and a correlation value between the output of the first adder and the determination error at the output of the first adder. Multiplied by the external tap weight controlled by The result is input as the first input of the second adder via the third delay element having a 1-symbol interval, and the output of the second adder is delayed by a 1-symbol interval by the fourth delay element. And a second recursive filter that feeds back the result of multiplication by the internal tap weight as the second input of the second adder, and the estimated value of the transmission path distortion is the external tap weight. Is controlled by combining the output of the multiplication result multiplied by with the output of the acyclic filter, and the internal tap weight is controlled by the correlation value between the output of the second adder and the decision error. A decision feedback equalizer characterized by the above.