On Mar 25, 9:35*am, "m" <stuttbr...@hotmail.com> wrote:
> >On Mar 25, 6:06=A0am, "m" <stuttbr...@hotmail.com> wrote:
> >> Hi everybody,
>
> >> I have a doubt that can be very obvious for you but it has driven me
> >> crazy. I want to know the performance of my demodulator by calculating
> th=
> >e
> >> BER curve respect to the EbN0. For that, I input some signals at
> differen=
> >t
> >> EbN0 but how should I calculate the EbN0? My method for a complex
> signal,
> >> as I read several times, is:
>
> >> Eb =3D S/R; where S is the signal power and R is the data rate
> >> N0 =3D N/Fs where N is the variance^2 and Fs is the sampling
> frequency.
>
> >> However, I've also seen that N0 =3D N/B where B is the bandwidth of
> the
> >> channel. Comparing my BER curve with the theoretical one, it is quite
> far
> >> and I would like to be sure that my EbN0 calculation is 100% right and
> th=
> >en
> >> check the demodulator.
>
> >> Could somebody please clear this point?
> >> Thank you very much
>
> >> Regards
> >> mary
>
> >Surprisingly, the explanation on the MathWorks' website is quite
> >good, and useful if you are simulating an AWGN channel in
> >discrete-time.
>
> >http://www.mathworks.com/support/sol....html?solution...
> >9I4S
>
> >Hope this helps,
> >Julius
>
> Thank you for your answer. Then it is using the sampling frequency.
> Regards
>
> Maria

The confusion comes in simulating a sampled representation with
oversampling. N0 should be the *density* of the AWGN. Hence
there is scaling depending on the sampling rate.

After matched filtering, they should all come out to be the same
as the "width" of the matched filter in Hz is set according to the
symbol rate, not the sampling rate.

Julius