Prev: Nonbinary or binary LDPC?
Next: Kalman Filter applied to Rangefinding Data
From: Dilip Warrier on 22 Jan 2010 09:01
On Jan 22, 6:01 am, "JAlbertoDJ" <nietoro...(a)yahoo.es> wrote:
> Hi,
>
> I need estimate Eb/No in a FSK of 8 symbols (8 tones).
>
> With FFT i get 8 values (8 bins), one for each tone.
>
> My doubt is how i can get first Es/No, and second Eb/No (I suposse Eb =
> Es/3  for 8FSK)
>
> I think that it could be, for example, in the case of received symbol is
> were symbol number 4:
>
> Es = Bin4^2
>
> No = Bin1^2 + Bin2^2 + Bin3^2 + Bin5^2 + Bin6^2 + Bin7^2 + Bin8^2
>
> Is it correct?
>
> Thank you

Yes, that is correct with some caveats:
1. This is a measure of the received Es/N0 (so, it includes the effect
of the channel) and not the transmitted Es/N0.
2. To get a reliable estimate, you will need to average over multiple
frames (where I define 1 frame to consist of the 8 bins you mentioned
above).
From: Jason on 22 Jan 2010 09:48
On Jan 22, 9:01 am, Dilip Warrier <dili...(a)yahoo.com> wrote:
> On Jan 22, 6:01 am, "JAlbertoDJ" <nietoro...(a)yahoo.es> wrote:
>
>
>
> > Hi,
>
> > I need estimate Eb/No in a FSK of 8 symbols (8 tones).
>
> > With FFT i get 8 values (8 bins), one for each tone.
>
> > My doubt is how i can get first Es/No, and second Eb/No (I suposse Eb =
> > Es/3  for 8FSK)
>
> > I think that it could be, for example, in the case of received symbol is
> > were symbol number 4:
>
> > Es = Bin4^2
>
> > No = Bin1^2 + Bin2^2 + Bin3^2 + Bin5^2 + Bin6^2 + Bin7^2 + Bin8^2
>
> > Is it correct?
>
> > Thank you
>
> Yes, that is correct with some caveats:
> 1. This is a measure of the received Es/N0 (so, it includes the effect
> of the channel) and not the transmitted Es/N0.
> 2. To get a reliable estimate, you will need to average over multiple
> frames (where I define 1 frame to consist of the 8 bins you mentioned
> above).

Another caveat: as your SNR decreases, the quality of your estimate
will decrease also. In your example, bin 4, which contains your signal
of interest, contains signal and noise. If the SNR is high enough, you
can consider the noise to have a negligible effect on the magnitude of
the value in that bin. If signal and noise are similar in power level,
however, your estimate will be affected accordingly.

Jason
From: JAlbertoDJ on 22 Jan 2010 09:57

>
>Yes, that is correct with some caveats:
>1. This is a measure of the received Es/N0 (so, it includes the effect
>of the channel) and not the transmitted Es/N0.


The measure of interest is about the signal received.

From: JAlbertoDJ on 22 Jan 2010 10:02
>
>Another caveat: as your SNR decreases, the quality of your estimate
>will decrease also. In your example, bin 4, which contains your signal
>of interest, contains signal and noise. If the SNR is high enough, you
>can consider the noise to have a negligible effect on the magnitude of
>the value in that bin. If signal and noise are similar in power level,
>however, your estimate will be affected accordingly.
>
>Jason
>

That's is true. Also, in the example we are suposse symbol received is
symbol number 4, but with a S/N very poor is probably that symbol received
were other different.





 | 
Pages: 1
Prev: Nonbinary or binary LDPC?
Next: Kalman Filter applied to Rangefinding Data