GMSK modulation and its PSD

Hi all
I'm working on GMSK modem simulation in MATLAB. I'm almost done with

the modulation part albeit i'm not sure if the model I used is
correct

adding to my trouble is the ambiguity with the Power Spectral Density

of the GMSK modulated signal (also the MSK modulated signal). Here
I'm

writing the algorithm I followed to do the simulation:

bitrate=4000bps BT=0.3

1.calculated the impulse response of the GaussianLPF with the formula

g(t)= convolution(gaussian pulse , rectangular pulse of width T) as

specified in the GSM 5.04 version 8.1.2 Release 1999.
In fact I directly used the impulse response specified in the

following link
>>http://www.emc.york.ac.uk/reports/linkpcp/appD.pdf

2. then I sampled the incoming binary data stream  exactly in the

middle of the bit period, then over sampled the stream with an over

sampling rate of 41
(here in a bit period first 20 samples will be 0 the 21st sample will

be +/-1 depending on the data thenext 20 will be 0 again)

3. I convolved these samples with the Impulse response I calculated

earlier.

4.then integrated the resulting signal to get 'y'

5.then calculated the I and Q channel streams as I=cos(y),Q=sin(y)

then modulated it with a I/Q modulator with Fc(carrier frequency=5000)

then I got the modulated signal as shown in the link below for the

data:
>>http://aycu19.webshots.com/image/45298/2003374167469833630_rs.jpg

Does a GMSK modulated signal look like that???
It shows constant amplitude and frequency deviations
the Ichannel vs Q channel (baseband) plot is as shown in the
following

link
>>http://aycu15.webshots.com/image/45574/2003383907486945009_rs.jpg

It is a circle with radius 1 as is expected of a GMSK signal


I calculated its PSD of the above signal using FFT and scaled its

frequency axis to 0 to Fs/2 (here Fs=16400Hz)
It is in the zoomed in figure(for viewability) shown in following link
http://aycu20.webshots.com/image/46459/2003353100129112203_rs.jpg

My doubt here is:
 In many text books and technical papers the PSD of GMSK signal shows

a peak at the carrier frequency but in my signal's PSD shows a dip

there I'm not getting where I went wrong.

PLEASE GUIDE ME WHERE I WENT WRONG AND PLEASE TELL ME THE CORRECT WAY

OUT OF THIS PROBLEM