# What is happening with the Cross-Correlation and Power Spectra of these signals?

Started by February 20, 2006
```Hi all,
I am trying to extract some caracteristic velocity from two time
signals obtained from two sources apart from a distance DX.
In order to get the velocity  I am dividing the distance DX by the
maximum time  lag of the cross-correlation. However I am getting the
maximum time lag to be at t = 0.
Does anybody have a clue about what is happening ?
My other doubt is when I try to get the Power Spectra (using the
periodogram function from Matlab) I get only one peak very close to
zero.
Does anybody have a clue about what is happening ?
My feeling is that is something wrong with the two signals (too short)
and the way I am using the periodogram function.
Any opinion is very welcome.
Thanks.

```
```desouzadacosta@pookmail.com writes:

> Hi all,
> I am trying to extract some caracteristic velocity from two time
> signals obtained from two sources apart from a distance DX.
> In order to get the velocity  I am dividing the distance DX by the
> maximum time  lag of the cross-correlation. However I am getting the
> maximum time lag to be at t = 0.
> Does anybody have a clue about what is happening ?
> My other doubt is when I try to get the Power Spectra (using the
> periodogram function from Matlab) I get only one peak very close to
> zero.
> Does anybody have a clue about what is happening ?
> My feeling is that is something wrong with the two signals (too short)
> and the way I am using the periodogram function.
> Any opinion is very welcome.
> Thanks.

Sounds like you might have a non-zero mean in either or both of the
signals. Try removing the mean before you run the cross correlation.
--
%  Randy Yates                  % "With time with what you've learned,
%% Fuquay-Varina, NC            %  they'll kiss the ground you walk
%%% 919-577-9882                %  upon."
%%%% <yates@ieee.org>           % '21st Century Man', *Time*, ELO
```
```Thanks,
Now the peak is very close to zero but not is zero.
I used the "xcov" function from matlab.

How about the power spectra ?
I am using the "periodogram" function but there is "psd" function too.
Any opinion is very welcome.
Alex

```
```desouzadacosta@pookmail.com wrote:
> Hi all,
> I am trying to extract some caracteristic velocity from two time
> signals obtained from two sources apart from a distance DX.
> In order to get the velocity  I am dividing the distance DX by the
> maximum time  lag of the cross-correlation.

Seems correct.

> However I am getting the
> maximum time lag to be at t = 0.
> Does anybody have a clue about what is happening ?

You may be using a signal that is close to stationary, like a sine.
This sort of thing might happen if you use a resonant transducer
and use an interval between repeated measurements that does
not allow the resonance to fade away between measurements.

> My other doubt is when I try to get the Power Spectra (using the
> periodogram function from Matlab) I get only one peak very close to
> zero.

Why do you use the power spectrum for this sort of estimate?
What is the relation between signal travel time and the power
spectrum?

> Does anybody have a clue about what is happening ?
> My feeling is that is something wrong with the two signals (too short)

It may very well be...

> and the way I am using the periodogram function.

I am almost certain about that...

> Any opinion is very welcome.
> Thanks.

Rune

```
```<desouzadacosta@pookmail.com> wrote in message
> Thanks,
> Now the peak is very close to zero but not is zero.
> I used the "xcov" function from matlab.
>
> How about the power spectra ?
> I am using the "periodogram" function but there is "psd" function too.
> Any opinion is very welcome.
> Alex
>
Periodogram is the discrete time estimate of continuous time PSD.

Tam

```
```Thank you for the answers.
My time series has 2900 elements. (too short ?)
I am trying to use the Power Spectra for having another estimate (not
for the velocity).
Does anybody have a clue what is the best way of getting a Power
Spectra in this case ?
I am trying to use  the matlab functions psd and periodogram

```
```desouzadacosta@pookmail.com wrote:
> Thank you for the answers.
> My time series has 2900 elements. (too short ?)
> I am trying to use the Power Spectra for having another estimate (not
> for the velocity).
> Does anybody have a clue what is the best way of getting a Power
> Spectra in this case ?
> I am trying to use  the matlab functions psd and periodogram

Now you are confusing me.

Exactly what are you trying to do? How do you measure

Rune

```
```Let me try to explain:
"Exactly what are you trying to do?": I have a time series with 2900
points.
I believe that using the power spectra I can extracting some useful
data such as
the dominant frequency.
"How do you measure your data?" My data is a array of lenght 2900. It
is extracted from a numerical simulation that is related to a wave
propagation phenomena.
"How do you process your data to extract information? "  I am not a DSP
expert and I am confused with the huge amount
of options I have to use with the matlab functions designed for
extracting the power spectra  (e.g. psd and periodogram).
So I need some methodologic guidelines to follow in order to extract
some useful information from my data.
Thanks.
Alex

```
```desouzadacosta@pookmail.com wrote:
> Let me try to explain:
> "Exactly what are you trying to do?": I have a time series with 2900
> points.
> I believe that using the power spectra I can extracting some useful
> data such as
> the dominant frequency.

That is true. The confusing issue is that you mentioned "velocity" in
your first post. It is not straight-forward to get from a power
spectrum
to a velocity estimate. It is possible -- I have done it -- but it does
require
a very specific measurement set-up along with some messy DSP.

> "How do you measure your data?" My data is a array of lenght 2900. It
> is extracted from a numerical simulation that is related to a wave
> propagation phenomena.

OK, a better question would have been "what is the measurement
geometry?" If you have one time series of a transient phenomenon,
you have one set of options. If you have an array measurement of
a stationary phenomenon, you have another set of options.

> "How do you process your data to extract information? "  I am not a DSP
> expert and I am confused with the huge amount
> of options I have to use with the matlab functions designed for
> extracting the power spectra  (e.g. psd and periodogram).
> So I need some methodologic guidelines to follow in order to extract
> some useful information from my data.

Which was why I asked the questions above. No one can help unless
you give every detail about the experiment or simulation you can
possibly
find. You wouldn't believe what inconspicuously-looking details decide
what can be done and what can not.

Rune

```