JPH0231524A - Adaptive equalizer - Google Patents

Abstract

PURPOSE:To converge a filter coefficient in a short time by respectively calculating evaluating functions to several sets of filter coefficients corresponding to an estimated length of a transmission line and selecting the set of the filter coefficients at which the evaluating function goes to the minimum and the reference waveform corresponding to the selected filter coefficient set. CONSTITUTION:An initial coefficient control circuit 11 calculates an evaluating function value expressed by the sum of an anti-jitter amplitude deviating quantity and inter-code interference quantity by observing the output of a digital filter 6 and selects the set of filter coefficients at which the evaluating function value can be made to the minimum from an initial coefficient memory 10 and, at the same time, selects the reference waveform corresponding to the selected filter coefficient set from a reference waveform memory 7a. Therefore, the anti-jitter amplitude deviating quantity and inter-code interference quantity can be suppressed irrespective of the line length and the filter coefficients can be converged in a short time.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an adaptive equalizer that highly accurately equalizes the frequency-loss characteristics of a signal transmission line by digital signal processing.

[Prior Art] FIG. 2 shows a schematic configuration diagram of a conventional adaptive equalizer. In the figure, (1) is the analog signal waveform input terminal IN, (2
) is a variable gain amplifier that keeps the amplitude of the signal waveform constant; (3
) is an A/D converter that samples and quantizes the output of the variable gain amplifier (2), and (4) is an A/D converter (3).
a gain control circuit (5) that monitors the output signal of the variable gain amplifier (2) and controls the gain of the variable gain amplifier (2) so that its amplitude is constant;
) is an adaptive coefficient memory that stores a set of filter coefficients for compensating line I5 characteristics, and (6) is a set of filter coefficients stored in the output of the A/D converter (3) and the adaptive coefficient memory. A digital filter that performs inner product calculation, (7
) is a reference waveform memory that stores a waveform that serves as a reference for the equalized output of the digital filter (6), and (8) is a reference waveform memory that observes the output signal of the digital filter (6) and stores it in the reference waveform memory (7). The adaptive coefficient memory (5
), and (9) is the output terminal OUT of the digital filter (6).

Next, the operation will be explained. During the training period, 3(T, +2nLT) = 1, a (To + (
2n-1) LT) = -1, a(To+IIIT) =
Q (m is a positive integer, m≠nL), nwo, ±1, ±2
A training pulse with a period of 2LT (T is the Symhol period) satisfying , . However, L is set sufficiently large so that the input waveform x (t) can be regarded as an isolated waveform. This input waveform x (t) is passed through a variable gain amplifier (2
) and then sampled by the A/D converter (3) at a frequency f'=4 times or more the symbol frequency f=1/T at -f=1/T'. A gain control circuit (4) controls the variable gain amplifier (2) so that the output signal of this A/D converter (3) has a constant amplitude. The output signal of the A/D converter (3) at this time is xb=
Represented by x (To + d + kT'), d>0.

Now, the digital filter (6) applies this sampled waveform xk% to the set of filter coefficients (C0, C1,
..., CN-1) and passes the output signal yk to the adaptive coefficient control circuit (8). The adaptive coefficient control circuit (8) uses sampled waveforms r, mi r (T(1+
d + kT'), (r(To + d + 2nLT
) = G, r(T,) + d + (2n-1)L
T) = G, r(To+ d + mT) = Ol
mTonL is a positive integer, n=o, ±1, ±2,...)
Take the difference ek = rk - yk with its output signal for
For sufficiently small Δ〉0, C''=C''-ΔekXk-jj=O11, ., N
-1 is performed, and the resulting set of filter coefficients is stored in the adaptive coefficient memory (5). Then, this operation is converted into the final evaluation function φ (co, -, CN-1) = λX, (+"+sk
-y+ak) 2+μX, y+ax ((ym+c
3'eik-1)2, (ymh-y+ai+++)2)
λ≧0, μ≧0, λ×μ=1, and m is a predetermined amount φ for 2L symbols. Continue until it becomes smaller.

The above φ is φ. When it becomes smaller, the training period ends and the data transmission period begins.

[Problem to be solved by the invention]

Since the conventional adaptive equalization amount is configured as described above, the same reference waveform must be used whether the line length is short or long, and therefore the amount of jitter amplitude deviation and code amount for the equalized output is This led to a deterioration in the amount of interference. Also, it took a long time for the filter coefficients to converge.

This invention was made to solve the above-mentioned problems, and it is possible to suppress jitter amplitude deviation and code amount interference regardless of the line length, and also to make the filter coefficients converge in a short time. The purpose is to obtain an adaptive equalizer.

[Means for Solving the Problems] The adaptive equalizer according to the present invention includes a first memory storing a plurality of filter coefficients each corresponding to the length of a transmission path, and a memory for storing filter coefficients as needed. and a third memory storing a reference waveform of the digital filter output, and a set of filter coefficients corresponding to the transmission path length estimated from the gain of the variable gain amplifier is stored in the first memory. At the same time, a reference waveform corresponding to this transmission path length is selected from the third memory, and the output signal of the digital filter is observed to determine the difference from the selected reference waveform. The minimum filter coefficient is stored in the second memory, and from then on, the filter coefficients stored in the second memory are used for digital filter calculations.

[Effect]

In this invention, an evaluation function is calculated for each set of several types of filter coefficients corresponding to the estimated transmission path, and a set of filter coefficients having the smallest value of the evaluation function among them is selected. Since the reference waveform is selected, it is possible to suppress the jitter amplitude deviation amount and the code amount interference amount regardless of the line length, and it is possible to converge the filter coefficients in a short time.

〔Example〕

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, the same reference numerals as in FIG. 2 indicate the same or corresponding parts, respectively. In addition to these, (7a) is a reference waveform memory that stores various reference waveforms corresponding to the length of the line, and (10) is a reference waveform memory that stores a set of filter coefficients for compensating for line characteristics. A read-only initial coefficient memory stored corresponding to the length (1
1) Observes the output signal of the digital filter (δ), calculates the evaluation function value expressed as the weighted sum of the jitter amplitude deviation amount and the code amount interference amount, and then calculates the evaluation function value so that it is the minimum value. A set of filter coefficients is selected from the initial coefficient memory, and the corresponding reference waveform is stored in the reference waveform memory (7a).
This is an initial coefficient control circuit that selects from among.

The operation of this embodiment will be explained below.

First, during the training period, a(T, +2nLT
) = 1, a(T0+ (2n-1)LT) =-1
, a(To+mT)xo (m is an integer, mv&nL), n
The transmission waveform is a training pulse with a period of 2LT (T is the symbol period) that satisfies =[], ±1, ±2,...
The transmission line output waveform x(t) is input manually from the input terminal (1). However, L is set sufficiently large so that the input waveform x (t) can be regarded as an isolated waveform. After amplifying this human-powered waveform x(t) to a constant amplitude with a variable gain amplifier (2), the A/D converter (3) converts it into a frequency f' = 4 times or more the symbol frequency f -17T, f = 1/ Sampled at T',
It is expressed by xk=x(T6+d+kT'), d>O. The gain control circuit (4) is connected to this A/D converter (3)
After controlling the variable gain amplifier (2) so that the output signal X, has a constant amplitude, - from its gain information, the line loss,
Therefore, the line length °C is estimated and this information is passed to the initial coefficient control circuit (11).

On the other hand, the initial coefficient memory (lO) has the length 11 J (01), j=0.1, which is obtained by increasing the line length in units of length v1,
..., a digital filter (6 )
0, l, ..., M-1), but the initial coefficient control circuit (11) stores the initial coefficient memory (10 ), N0 filter coefficients with corresponding line lengths longer than 1 and N1 filter coefficients with corresponding line lengths shorter than l are selected from among the sets of filter coefficients as described above. In other words, there is a j. (0≦jo≦
For M-1), when U, , = X, I JO-N
,...,IJ,-1,It JOx 1 jo+
1. Select a set of N0+N+1 filter coefficients for the line length I-JO+No. Also, in the reference waveform memory (7a), the length V and the length u which is increased by the line length in units are stored in the reference waveform memory (7a). (m), j=0.1, ·, for the line length of M-1, 'f'i = T 6 + d + d
'7.

d'J> 0, j=0.1,...-1M-1, however, 1≦s J< ao -co (j (co, n
=0. Waveform rk(Sj) = r(sj: "f" + kT sampled at frequency T° for ±1, ±2,...
'), and for 1≦S<C0, the evaluation function φ(s)=min(λΣ(r+a+c (S) −y+
am) ”μΣa+ax ((ymk-ymm-+)
2, (ymk-ymk+1)2): V va k”Σ
CIXmk-J, co((, When (cxz) is considered as a function of S, φ(SJ) −win (φ(s); 1≦S<00
, j = Q, 1, . . , M-1) are stored. The initial coefficient control circuit (11) inputs N, +N, +1 reference waveforms rk (SJO-Nl), "", r 1 (Sto
-1), rm(S'o), rk(Sjo++), -, r
Select k(Sro”No).

Next, the initial coefficient control circuit (11) j = Jo−N+,
..., jo-1, Jo, jo"l, ", jo"No when using the set of filter coefficients ~ 1 (C"; i = Q, 1, ..., N-1) Output signal yk, J (k
= 0, ±1, ±2, ...) and evaluate the evaluation function value ~ I φJ (CJ) = λΣ(r+am (SJ) -Ym
k, J) 2+μΣmaX CCyr6−”−Ym
k-small, j) 2, (3'mk, J-Ymk+o, J
) ”). Then, “’ , Jo−1, jo, jo”l, ..., jo
. . , CJln) and the reference waveform "k(SJ+m1n)" are selected.

Now, the adaptive coefficient control circuit (8) uses a set of filter coefficients (C0, Cl...,
cN-1) and the input waveform xk, the reference waveform '=
Difference from rk(S"++un) ek=rk-3'k
For a sufficiently small Δ〉0, Cl=Cl−ΔekX+c−Ii +=0.1,・,N
-1 is performed and the resulting set of filter coefficients is stored in the adaptive coefficient memory (5). Then, this operation is performed using the final evaluation function φ = (C0,...cN-1)mi λ Σ (r
+mk-y+mk)"μΣmaX ((ymk, j-
y+mk-+1. )2, (y+ak, J-Vmm**
, J) ”) is smaller than the predetermined amount φ.

The above φ is φ. When it becomes smaller, the training period ends and the data transmission period begins. During the data transmission period, the filter coefficients stored in the adaptive coefficient memory (5) are always used.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, an evaluation function is calculated for each set of several types of filter coefficients corresponding to the estimated transmission line length, and among them, the value of the evaluation function is By selecting the smallest set of filter coefficients and the corresponding reference waveform, it is possible to suppress jitter amplitude deviation and code amount interference, regardless of line length, and reduce the number of filters. This has the effect of being able to converge within a certain amount of time.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.
FIG. 2 is a block diagram showing the configuration of a conventional adaptive equalization amount. (2)...Variable gain amplifier, (3)...-A/D converter, (4)...Gain control circuit, (5)...Adaptive coefficient memory, (6)...Digital filter , (7
a)...Memory for reference waveform, (8)...Adaptive coefficient control circuit, (lO)...Memory for initial coefficient, (11)...
...Initial coefficient control circuit. In each figure, the same reference numerals indicate the same or corresponding parts. Figure 1

Claims (1)

[Claims] A variable gain amplifier that amplifies the signal sent from the transmission line, an A/D converter that samples and quantizes the output of this variable gain amplifier, and an A/D converter that samples and quantizes the output of this A/D converter so that the amplitude of the output signal of this A/D converter is constant. A first control circuit that controls the gain of the variable gain amplifier receives the output of the A/D converter and performs an inner product calculation with a set of filter coefficients for compensating the frequency attenuation characteristic of the transmission line. A digital filter to be output, a first memory storing the filter coefficients corresponding to the length of the transmission path, a second memory storing the filter coefficients as needed, and a reference waveform of the output of the digital filter. a third memory storing a set of filter coefficients corresponding to the length of the transmission path, and a third memory storing a set of several types of filter coefficients corresponding to the length of the transmission path estimated from the gain information of the variable gain amplifier. a second control circuit that selects a reference waveform from the third memory and provides it to the calculation of the digital filter, and selects a reference waveform corresponding to these transmission path lengths from the third memory, and observes the output signal of the digital filter; a third control circuit that stores in the second memory a filter coefficient that has a minimum difference from the reference waveform selected by the second control circuit, the filter coefficients stored in the second memory; An adaptive equalizer characterized by transmitting data using.