Preamble correlator, Simulink problem

[Antonio_Magma]

xcorr simulink

Season's greetings!

I'm currently trying to build a preamble detector for my OFDM
transceiver which i have built in Simulink for the purpose of frame
synchronization. This is kinda a lenghty post, but i hope
those who can help me please take some time and take a look at it.
Appreciate it!

This is what i have done so far. This is autocorrelation of a PN
sequence:
http://geocities.com/antonio_magma/corr/corrawgn.gif
I used 64 IFFT to convert the binary bits into real-valued time domain
(with conjugate symmetry), pass through AWGN and then correlated. This
is the resulting graph after autocorrelation:
http://geocities.com/antonio_magma/corr/corrawgngraph.gif
It looks fine with the first sample having a larger magnitude as
compared to the others.

Next, i've replaced autocorr with cross-corr just to observe the
differences:
http://geocities.com/antonio_magma/corr/xcorrawgn.gif
At the 2nd input of the xcorr i've connected it to the IFFT output to
act as a reference for the received signal after AWGN. And the result i
got is this:
http://geocities.com/antonio_magma/corr/xcorrawgngraph.gif

1st question:
Why is there such a huge difference with the results? In the 2nd graph
i'm getting almost noise like signal. Does it mean i should use
autocorr instead of xcorr with my model? Or am i building the model
wrong?

Next, because typically the preamble is repeated and transmitted
multiple times for synchronization, i've tried to concatenate the
output of IFFT forming 4 identical sequences to be transmitted and
correlated.

This is what i built:
http://geocities.com/antonio_magma/corr/4corrawgn.gif
And the result is:
http://geocities.com/antonio_magma/corr/4corrawgngraph.gif

And for the xcorr:
http://geocities.com/antonio_magma/corr/4xcorrawgn.gif
http://geocities.com/antonio_magma/corr/4xcorrawgngraph.gif

2nd question:
With reference to the 4corrawgngraph, the 1st peak became even larger
which i guess could be due to the superposition of the 4 sequences. But
shouldnt the output be like this?
http://geocities.com/antonio_magma/corr/corrop.gif

My final objective is to build a correlator which produces and output
like the 1 above and this:
http://geocities.com/antonio_magma/corr/yunchiu-corr.gif

I understand the basic theory behind double correlation but how do i
implement it to obtain an output like yunchiu-corr? Guys, is it
something that i've done wrong with my simulation (which i think is the
case) or i should have designed it differently, any ideas please let me
know coz i've been stuck at this for quite some time. Thx guys...