GMSK Question

Hi,

I am trying to simulate the GSM physical layer and have a simple question
regarding GMSK modulation:

I am trying to implement the GMSK modulator using the quadrature baseband
method outlined in: http://www.emc.york.ac.uk/reports/linkpcp/appD.pdf
which states : NRZ sequence -> Gaussian filter convolution -> Integration
-> I/Q decomposition -> Modulation. 

However, i have checked several other posts on this forum such as:
http://www.dsprelated.com/showmessage/45749/1.php  
and people say that before integrating, I should express the convolved
sequence as a phase rotation in radians and afterwards raise to the
exp(j*phi) to get the complex I & Q signals. 

I doubt that these two methods give out the same result. Which one should
be used for a GSM simulation ?

Thank you in advance,
John

On Feb 13, 12:29 pm, "ik303" <john.kon...@gmail.com> wrote:
> Hi,
>
> I am trying to simulate the GSM physical layer and have a simple question
> regarding GMSK modulation:
>
> I am trying to implement the GMSK modulator using the quadrature baseband
> method outlined in:http://www.emc.york.ac.uk/reports/linkpcp/appD.pdf
> which states : NRZ sequence -> Gaussian filter convolution -> Integration
> -> I/Q decomposition -> Modulation.
>
> However, i have checked several other posts on this forum such as:http://www.dsprelated.com/showmessage/45749/1.php
> and people say that before integrating, I should express the convolved
> sequence as a phase rotation in radians and afterwards raise to the
> exp(j*phi) to get the complex I & Q signals.
>
> I doubt that these two methods give out the same result. Which one should
> be used for a GSM simulation ?
>
> Thank you in advance,
> John

The two references you cite describe exactly the same process. All
they're referring to with "expressing the sequence as a phase
rotation" is scaling the Gaussian filter output (or the coefficients
of the filter itself) such that when you pass them through your
integrator, a single Gaussian pulse will integrate to a value of pi/2.
You then take the cosine and sine of these quantities to get your I
and Q channels, then you go on to the modulator.

So, either way will work.

Jason

Thanks for the reply.

John

Hi,

I'm trying to build a GMSK modulator following the procedure described in
the aforementioned docs.
Do I have to remove the gaussian filter delay after convolution?
If so, do I have to pad with zeros the input signal?
Thanks

Alberto