FPGA Central

Hello friends ,
I need to develop a system that given two
signals
x(n)=c(n)+f(n)
y(n)=A.f(n+t)

I could estimate what is the amplitud constant "A" and the phase
shifting "t" of signal y(n) comparing with f(n) of x(n).
I know this is statistical processing and I havent develop much about
it so I need a guideline to start with. I have a couple of books wich
I am reading but I think I am going to slow. I know that cross-
correlation has something to do with this as well, but still I cant
get any conclusions yet.

Hope anyone here can help me

Alfredo

Alfredo wrote:
> Hello friends ,
> * * * * * * * * * * * I need to develop a system that given two
> signals
> x(n)=c(n)+f(n)
> y(n)=A.f(n+t)
>
> I could estimate what is the amplitud constant "A" and the phase
> shifting "t" of signal y(n) comparing with f(n) of x(n).

You have x(n) and y(n), right? What kind of signals are c(n) and f(n)?
What do you know about c(n) and f(n)? Can you specify x(n) and then
measure y(n)?

> I know this is statistical processing and I havent develop much about
> it so I need a guideline to start with. I have a couple of books wich
> I am reading but I think I am going to slow. I know that cross-
> correlation has something to do with this as well, but still I cant
> get any conclusions yet.

You can write your problem as a Wiener filter in System Identification
configuration. If c(n) is what I think it is (white noise), this would
be a feasible solution.

Regards,
Andor

>You can write your problem as a Wiener filter in System Identification
>configuration. If c(n) is what I think it is (white noise), this would
>be a feasible solution.
>
>Regards,
>Andor
>
*****************************
If i try to estimate t and A from x and y, via Wiener filter, the
knowledge of the cross-correlation of A with x and y and of t with x and
is required, right? So how can the Wiener filter be a feasible solution?
(the question assumes that A and t are constants)

Manolis

On 27 Mrz., 10:32, "Manolis C. Tsakiris" <el01...@mail.ntua.gr> wrote:
> >You can write your problem as a Wiener filter in System Identification
> >configuration. If c(n) is what I think it is (white noise), this would
> >be a feasible solution.
>
> >Regards,
> >Andor
>
> *****************************
> If i try to estimate t and A from x and y, via Wiener filter, then
> knowledge of the cross-correlation of A with x and y and of t with x and y
> is required, right? So how can the Wiener filter be a feasible solution?
> (the question assumes that A and t are constants)

Not quite sure what you mean by the cross-correlation of "A with x and
y", since A is a constant. You can write Alfredo's equation in the
form

y(n) = A x(n+t) + c'(n)

with c'(n) = - A c(n+t). If c(n) (and therefore c'(n)) is white noise,
this is a classical system identification setup. x(n) is the input
into the Wiener filter and y(n) is the desired output.

Regards,
Andor


>
> Manolis

OK, thanks in advance guys.
his project is for a RF repeater with Interference Cancellation.

x(n) is a signal in a reception antenna composed by a CDMA carrier
"c(n)" and a feedback interference from the transmitting antenna
"f(n)".
So x(n) is what I receive in the reception antenna.

y(n) is the signal that is going to be transmitted and it is delayed
enough time so that "f(n)" from x(n) arrives at the same time.

I think I made a mistake in my last post. I think that this system
would be better expressed by....

x(n)=c(n)+Af(n-t)
(Received signal= actual reception+feedback interference with Amplitud
and phase change)

The other signal I have it is f(n) before transmitting

y(n)=f(n)

I hope I made it clearer

cheers
Alfredo





On 27 mar, 06:48, Andor <andor.bari...@gmail.com> wrote:
> On 27 Mrz., 10:32, "Manolis C. Tsakiris" <el01...@mail.ntua.gr> wrote:
>
> > >You can write your problem as a Wiener filter in System Identification
> > >configuration. If c(n) is what I think it is (white noise), this would
> > >be a feasible solution.
>
> > >Regards,
> > >Andor
>
> > *****************************
> > If i try to estimate t and A from x and y, via Wiener filter, then
> > knowledge of the cross-correlation of A with x and y and of t with x and y
> > is required, right? So how can the Wiener filter be a feasible solution?
> > (the question assumes that A and t are constants)
>
> Not quite sure what you mean by the cross-correlation of "A with x and
> y", since A is a constant. You can write Alfredo's equation in the
> form
>
> y(n) = A x(n+t) + c'(n)
>
> with c'(n) = - A c(n+t). If c(n) (and therefore c'(n)) is white noise,
> this is a classical system identification setup. x(n) is the input
> into the Wiener filter and y(n) is the desired output.
>
> Regards,
> Andor
>
>
>
> > Manolis

Alfredo wrote:
> OK, thanks in advance guys.
> his project is for a RF repeater with Interference Cancellation.
>
> x(n) is a signal *in a reception antenna composed by a CDMA carrier
> "c(n)" and a feedback interference from the transmitting antenna
> "f(n)".

What is a "CDMA carrier c(n)"?

> So x(n) is what I receive in the reception antenna.
>
> y(n) is the signal that is going to be transmitted and it is delayed
> enough time so that "f(n)" from x(n) arrives at the same time.
>
> I think I made a mistake in my last post. I think that this system
> would be better expressed by....
>
> x(n)=c(n)+Af(n-t)
> (Received signal= actual reception+feedback interference with Amplitud
> and phase change)
>
> The other signal I have it is f(n) before transmitting
>
> y(n)=f(n)
>
> I hope I made it clearer

So you have f(n) and x(n), and you want to estimate A and t in the
equation

x(n) = c(n)+Af(n-t)?

I guess c(n) isn't white noise as I initially assumed?

Regards,
Andor

On 27 mar, 08:39, Andor <andor.bari...@gmail.com> wrote:
> Alfredo wrote:
> > OK, thanks in advance guys.
> > his project is for a RF repeater with Interference Cancellation.
>
> > x(n) is a signal in a reception antenna composed by a CDMA carrier
> > "c(n)" and a feedback interference from the transmitting antenna
> > "f(n)".
>
> What is a "CDMA carrier c(n)"?
>
> > So x(n) is what I receive in the reception antenna.
>
> > y(n) is the signal that is going to be transmitted and it is delayed
> > enough time so that "f(n)" from x(n) arrives at the same time.
>
> > I think I made a mistake in my last post. I think that this system
> > would be better expressed by....
>
> > x(n)=c(n)+Af(n-t)
> > (Received signal= actual reception+feedback interference with Amplitud
> > and phase change)
>
> > The other signal I have it is f(n) before transmitting
>
> > y(n)=f(n)
>
> > I hope I made it clearer
>
> So you have f(n) and x(n), and you want to estimate A and t in the
> equation
>
> x(n) = c(n)+Af(n-t)?
>
> I guess c(n) isn't white noise as I initially assumed?
>
> Regards,
> Andor


CDMA Carrier = Type of signal in communications.
But for the purpose of the system we can consider c(n) as an unwanted
signal. The overall system should be able to get c(n) clear of f(n) by
having the right amplitud and phase shift.

You are right I have x(n) and f(n) and I want A(amplitud) and
t(phase).

I guess c(n) isn't properly white noise but is an unwanted signal so
maybe we could consider it as noise.

Cheers
Alfredo

(BTW , Andor do you have any email where I can chat with you about
this ?)