Time delay - Cross Correlation Situation

Hi everyone...

I'm trying to calculate time delay using cross correlation but i don't know why i'm having this problem:

a=rand(1,10); %random excitation

x1=[a zeros(1,90)]; %Input signal

x2=[zeros(1,90) 0.8.*a]; %Output signal is a delayed version of x1 attenuated

Delay is equal to 90? right?

Well i should be able to calculate this delay from xc=xcorr(x1,x2);

but once i do this the peak is at xc(10) instead of be positioned at xc(190) now when i do xc=xcorr(x2,x1) peak is right positioned.

I don't know if I made myself clear, but i think that xc=xcorr(x1,x2) should be the correct way to do it....

Please help me here...

Answer 1

[Honglei Chen]

The order basically determines which one is the reference, but you need to set your lag right. The delay between the two is basically 90, so one will give you 90 and the other one is -90. Try the following:

[xc,lags] = xcorr(x2,x1);
plot(lags,xc)
[xc,lags] = xcorr(x1,x2);
plot(lags,xc)

[Gabriel González]

I understand, but still i'm not sure because if i do [xc,lags] = xcorr(x1,x2) i should have a delay equal to 90 not -90 which is what i got...

could I be conceptually wrong?

[Honglei Chen]

You may be using a different convention? The definition of the correlation in the doc page may clarify it for you.

http://www.mathworks.com/help/toolbox/signal/ref/xcorr.html