Least mean square to equalize optical fiber channel

Question

I used a Matlab code of LMS (least mean square algorithm) to equalize the effect of the channel, it is working for a tapped delay channel generated in MATLAB but for optical fiber channel using optisystem program, it doesn't work well, i think the problem is in optical fiber impulse response, as it is not finite impulse response and the algorithm needs a finite impulse response but how to solve this problem?

the constellation of MATLAB channel and optical fiber channel, it is clear that the second constellation hasn't been equalized well.

constellation:

Here is an example of a noisy channel where LMS algorithm couldn't compensate its effect, is there any way to enhance its performance at this SNR :

% // Set up parameters and signals.

M = 4; % // Alphabet size for modulation

msg = randint(15000,1,M); % // Random message


modmsg = pskmod(msg,M); % // Modulate using QPSK.


trainlen = 5000; % // Length of training sequence


chan = [.986; .845; .237; .123+.31i]; % // Channel coefficients


filtmsg = filter(chan,1,modmsg); % // Introduce channel distortion.

filtmsg=awgn(filtmsg,10,'measured');

% // Equalize the received signal.

eq1 = lineareq(8, lms(0.01)); % // Create an equalizer object.

eq1.SigConst = pskmod([0:M-1],M); % // Set signal constellation.

[symbolest,yd] = equalize(eq1,filtmsg,modmsg(1:trainlen)); % // Equalize.

% // Plot signals.

h = scatterplot(filtmsg,1,trainlen,'bx'); hold on;

scatterplot(symbolest,1,trainlen,'g.',h);

Answer 1

Maybe you should have more samples. The "averaging" effect of the LMS doesnt work when the spectrum is close to the Nyquist criteria.

Other problem could be that the noise is not white (wide-band) enough. So the noise-samples are coherent.