JPH01188117A - Transversal type automatic equalizer - Google Patents

Abstract

PURPOSE:To simultaneously ensure both the long time stability of tap gain correction control and the sufficient equalizing performance by setting the leakage object of a tap in the vicinity of the main tap based on the calculation from the input waveform. CONSTITUTION:As to a tap in the vicinity of the main tap, the leakage object is set based on equation and other taps are set to 0, where c' is leak object vector (c'-M,...,c'O,...,c'N), A is an input signal matrix (M+N+1)-order square matrix, elements (p, q) of the matrix A (-M<p, q<N) are expressed in equation I, v is a channel vector (M+N+1)-order), the p-th element of (v) is expressed in equation II, Xk, i is an input signal (a sample at time kT of the i-th tap), dK is a desired signal (sample at time kT), T is a transversal filter output sampling period, and [P, Q] are range of input signal matrix correlation calculation. Thus, both the stability and the equalizing performance of the tap gain correction control are ensured.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a transversal automatic equalizer that sequentially corrects and controls tap gain.

(Prior Art) The tap gain correction algorithm of a conventional transversal automatic equalizer that sequentially corrects and controls the tap gain can be expressed as follows.

at(n+1)=ct(n)−αxist(yh−dJ
t x=-Mm...+N■ However, Ct(n): i-th tap gain after first correction Xk+l
: Input signal (i-th tap output at sampling time kT) yk : Output signal (transversal filter output at sampling time kT) dk : Desired signal (sampling time kT) α
:Positive infinitesimal constant By the way, the tap gain correction control of the transversal automatic equalizer as described above is stable in many cases, but depending on the conditions it is not necessarily stable and the tap gain correction control continues for a long time. In this case, a phenomenon may be observed in which the tap gain exceeds its original desired value and its absolute value continues to increase indefinitely. This happens because 1/2 T'> f a, aX exists between the tap interval T' of the 1-Rance versal filter and the highest frequency component maX of the input signal.
This is a case where there is the following relationship and the tap gain is a digital representation with a finite word length. (References: R, D
.

Gitlin et al., The Tap-Lea
kage Algo-rithm: An A1g
orith+m for the 5table 0p
era-tion of a digital
y Implemented, Fraction
--nally 5paced Adap
tive Equalizer, B, S.

T, J, vol, 61. No, 8. pp, 1817-
1839. Oct.

■Equation is 1-pass band of Lanceversal filter (1/
2T') means that there is still a surplus even if the signal band is completely covered, but this relationship only holds true when using an oversampling equalizer in synchronous data transmission or a goose 1-canceller. This is the case for equalizers that target arbitrary waveforms such as , and is a field of equalizers that is important in practice.

It is also well known that in order to eliminate such long-term instability of tap gain modification control, it is effective to add leakage to tap gain modification control (see the above reference). In other words, instead of the formula α), ct(n+1)=(1-γ)at(n) ayc**i
cy* dJ ■or at(n+1)=ct(n) y°5q(ct(n))
-αxk*t(yz-dt) (IA, where γ: Modify the tap gain by the leakage constant (positive and sufficiently smaller than 1). In this way, all tap gains have a centripetal force toward O. works, it is possible to prevent excessive growth of tap gain.

Although this avoids long-term instability in tap gain modification control, it is also true that the performance as an equalizer is sacrificed. This is because the convergence value of the tap gain in this method deviates to be somewhat smaller (in absolute value) than the original optimal value. In order to minimize this sacrifice, for the main tap (the tap with the maximum tap gain), instead of giving leakage towards 0, it is also possible to give leakage towards some suitable non-O value (for example 1). It is being done.

However, even with such a method, it may be difficult to simultaneously ensure both long-term stability of tap gain modification control and sufficient equalization performance depending on the transmission path characteristics. This is because when looking at the optimal tap gain after convergence in the vicinity of the main taps, there is not just one clearly prominent main tap, but multiple taps come together to form a main tap group. This is the case. At this time, if the leakage target value is set to 1 for only one specific tap, the tap gain error near the main tap will be large even after the tap gain converges, and sufficient equalization performance will not be obtained.

(Problems to be Solved by the Invention) As described above, conventionally known leakage control methods simultaneously ensure both long-term stability of tap gain modification control and sufficient equalization performance depending on the transmission path characteristics. The disadvantage was that it was difficult to do so. The present invention aims to provide a transversal type automatic equalizer that solves this drawback by setting the leakage target value of the tap near the main tap based on calculation from the input waveform. be.

[Structure of the invention]

(Means for Solving the Problems) In order to achieve the above object, in the present invention, in a transversal type automatic equalizer in which leakage is added to the tap gain correction control, the leakage target value is set in the vicinity of the main tap. For tap, the equation: % formula % However.

c': Leak target value vector (C-He...+ Q Op...e Q N) A
: Input signal matrix ((M+N+1)) order square matrix) A's (
p, q) element: (where -M≦p, q≦N) apq”
ΣX k + P X k + Qk! P V Two-channel vector (order (M+N+1)) p-th element: vp=ΣXksPdkk! P Xkti: Input signal (sample value at time kT of i-th tap) dk: Desired signal (sample value at time kT) T Nidoran universal filter output sampling period [P, Q]: Based on input signal matrix correlation calculation range setting, and other taps are set to O.

(effect) In this way, the section [-M] near the main tap.

N], the tap gain does not leak toward O, but instead leaks toward c, H''aN.

In other intervals, the tap gain leaks toward zero. Therefore, it is possible to realize a transversal automatic equalizer that can sufficiently suppress overgrowth of tap gain while at the same time minimizing deterioration in equalization performance.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing one configuration of the present invention.

This configuration attempts to make the transmission characteristics of the entire transmission system including the transversal equalizer into a desired shape by controlling the tap gain modification of the transversal equalizer using a known signal waveform as a training signal. It is something.

In this embodiment, the sampling period T and 1
~The tap interval T' of the Lanceversal filter is equal. Further, the value of T is selected so as to satisfy Equation 0.

On the transmitting side, a uniform random number sequence (rk) with an amplitude in the range of -0.5 to 0.5 is sent out at a clock cycle T. This signal enters the receiver input terminal 1 via a transmission path and is sampled at a clock period T by a sampler 2. The input signal (X,), the result of which is the input signal (X,), is transferred to a shift register 4. with a clock period T. Tap gain multiplication circuit 5
．． The signal is outputted to an output terminal 40 through a transversal filter 3 consisting of an adder circuit 6 and an adder circuit 6. The simultaneous coefficients, ie, the tap gains, given to the tap gain multiplication circuit 5 are stored in the tap gain memory 7, and are successively modified by a method described later. A reference signal memory 8 on the receiving side stores a copy of the transmission training signal (r,), and by passing this through a reference transmission circuit 9, a desired signal (dk) is obtained.

On the other hand, in a predetermined section (xp-~XQ of the input signal (xk)
+N) is once taken into the input signal memory 13, and the arithmetic circuit 14 uses this and the output (dk) of the reference transmission circuit 9.
From this, a−H˜ON is calculated using the formula 0. In addition,
Here, since T'=T, X k * i is X k-
It can be replaced with i. Q − found here
H-(', H is stored in the leak target value memory 15.

Also, the same signal (yk) that is output to the output terminal 40
is led to the subtraction circuit 10, where the desired signal (dk) from the reference transmission circuit 9 is subtracted, and the error signal (e,
). This error signal is supplied in parallel to a multiplication circuit 11 attached to each tap, and the product of each tap with the output (No. ii Xk*i) is calculated.

This product is information for tap gain modification, and the tap gain modification circuit 12 uses this to modify the tap gain according to the following equation.

Here, Q-H''''Cp4 is the value previously stored in the leakage target value memory 15. (This formula is
This is a modification of equation (3), which is a conventional leak control method.

In addition, modifications of formula 1 (0) also correspond to the spirit of the present invention, but their description is omitted to avoid complexity. ) If tap gain modification control is performed as described above, it will not only have the effect of preventing excessive tap gain growth, but also make the tap gain convergence value near the main tap extremely close to the optimal value, similar to conventional leakage control. A new effect arises. This eliminates the tap gain error near the main tap due to leakage, which was a problem with the conventional method, and makes it easier to obtain the desired equalization performance.

Note that, in general, transmission path characteristics do not change over time. Assuming that a known signal is sent intermittently from time to time on the receiving side as a transmission signal, the input signal at that time is taken into the input signal memory 13, c-1~ON is determined by the operation 20, and leakage is detected. The contents of the target value memory 15 are updated. At this time, the contents of the leak target value memory 15 may be completely replaced with the new value, or may be replaced with an appropriate linear sum of the old value and the new value. The latter cannot cope with rapid changes, but has the advantage of being able to reduce the effects of noise.

Further, although this embodiment shows the case where a uniform random number sequence is used as the training signal, the same objective of equalizing the transmission path characteristics can also be achieved by using a periodic impulse waveform as the training signal. One example is a ghost cancel race using a differential waveform of a vertical synchronization signal as a training signal.

It is clear that the present invention can also be applied to such Gose 1 to Canceller.

FIG. 2 shows another embodiment of the present invention. This is the case with an oversampling equalizer in baseband multilevel data transmission. That is, for a clock period T that corresponds to the signal transmission rate.

The clock period T' of the transversal filter 3 is T
'< T (often used is T' =
T/2).

Since it is very similar to the configuration shown in Figure 1, only the differences will be described. First, the transmitted 1M signal is a data sequence subjected to N-value amplitude modulation. Next, the clock period of the sampler 2 and the transversal filter 3 is not T but T'. moreover,
The output of the transversal filter 3 is led to an amplitude determination circuit 21 via a sampler 20 with a sampling period T. The output corresponds to the output of the reference transmission circuit 9 in the configuration shown in FIG. That is, in the subtraction circuit 10, the error signal (ek) is obtained by subtracting the output of the amplitude determination circuit 21 from the output of the sampler 2. This and each tap output of the transversal filter 3 are introduced into the multiplication circuit 11,
The process in which the tap gain modification calculation of 20 is performed in the tap gain modification circuit 12 using the product as tap gain modification information is the same as in the case of FIG.

The first step is to take in a part of the received signal into the received signal memory 13.
The case is the same as the one shown in the figure, but in the subsequent calculation to obtain the glue mark value, the location where the desired signal (dm) is obtained is different, and in this configuration, the output of the amplitude determination circuit 21 is
Used as k).

In addition, in the explanation of the above two embodiments, the transversal filter 3, the multiplication circuit 11, etc. were explained as if they were separate hardware, but this is for convenience of explanation and does not necessarily have to be so. Of course, it can also be realized by a microprocessor, DSP (digital signal processing LSI), etc.

Furthermore, in the above two embodiments, the signals such as (X, ) and the Lanceversal filters 1 to 3 are one-dimensional, but the purpose of the present invention can be achieved even if these are expanded to two-dimensional in-phase and orthogonal components. Do not deviate from this. At this time, Xk+ dk+ of formula 0
It is sufficient to regard C, etc. as a complex number.

〔Effect of the invention〕

According to the present invention, in a 1-rance versal type automatic equalizer, near the main tap where the absolute value of the tap gain at the time of convergence is expected to be relatively large.

The leakage target value was set near the original convergence value by calculation based on the input waveform, and for remote taps where the absolute value of the tap gain at the time of convergence was expected to be relatively small, the leakage target value was set to 0. Since leakage control is performed on top of the system, it is possible to achieve a transversal automatic equalizer that sufficiently achieves the original purpose of leakage control, which is to suppress tap gain overgrowth, while at the same time minimizing the deterioration of equalization performance. can do.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing one embodiment of the invention, and FIG. 2 is a diagram showing another embodiment of the invention. 3...Transversal filter 7...Tap gain/l) Memory 12...Tap gain correction circuit 13...Input waveform memory 14...Arithmetic circuit 15
...Leak target value memory agent Patent attorney Nori Chika Yudo Mitsuyuki Matsuyama

Claims (1)

[Claims] In a transversal automatic equalizer in which leakage is added to tap gain correction control, the leakage target value for taps near the main tap is expressed by the equation: c'=A^-^1v, where: c ': Leak target value vector (C'_-_M,..., c'O,..., c'_N)
A: Input signal matrix ((M+N+1)) order square matrix) (p, q) elements of A: (However, -M≦p, q≦N) ▲ Contains mathematical formulas, chemical formulas, tables, etc. ▼ v: Channel vector ( (M+N+1)th) pth of v
Elements: ▲There are mathematical formulas, chemical formulas, tables, etc.▼ X_k_,_i: Input signal (sample value at time kT of the i-th tap) d_k: Desired signal (sample value at time kT) T:
Transversal filter output sampling period [P, Q]: A transversal type automatic equalizer characterized in that it is set based on the input signal matrix correlation calculation range, and other taps are set to 0.