Forums

Min. sampling rate required to resolve 0.1ms delay using cross correlation techniques

Started by Unknown March 13, 2005

Hi all,
I have a question for those who are familar with the 'cross correlation' function in Matlab. If I wanted to resolve a 0.1ms delay between two sampled audio signals using the cross correlation function in Matlab, and if I were to use 20kHz speech signals sampled at rate of say 50KHz (to satisfy the Nyquist criterion) would this be a sufficient to resolve a 0.1ms delay between the two signals? If I use the logic that a 0.1ms period corresponds to a 10KHz 'frequency', I thought probably be ok. But is there a general rule for this sort of thing? Would I be better of using a much higher sampling rate (say in the 100Kz, or 1MHz range?)

I am relatively inexperienced with particular feature of Matlab, so thank you in advance for any for any help!
Alicia




Alicia,

since your sampling rate is 50kHz, that means the time between two
samples in your signals is 20us. So when you do xcorr, the
cross-correlation can detect lags greater than 20us. Since you are
trying to detect 100us (= 0.1ms), you should be fine.

If you go for even high sampling rates, the granularity will be finer.
e.g. if instead of 0.1ms, your delay is say 0.11ms. Then with 50kHz
sampling you won't be able to distinguish between 0.1 and 0.11ms, but
if you go to a higher sampling rate you will be.

hope this helps
Nandan --- In matlab@matl..., not_that_kind6@y... wrote:
>
> Hi all,
> I have a question for those who are familar with the 'cross
correlation' function in Matlab. If I wanted to resolve a 0.1ms delay
between two sampled audio signals using the cross correlation function
in Matlab, and if I were to use 20kHz speech signals sampled at rate
of say 50KHz (to satisfy the Nyquist criterion) would this be a
sufficient to resolve a 0.1ms delay between the two signals? If I use
the logic that a 0.1ms period corresponds to a 10KHz 'frequency', I
thought probably be ok. But is there a general rule for this sort of
thing? Would I be better of using a much higher sampling rate (say in
the 100Kz, or 1MHz range?)
>
> I am relatively inexperienced with particular feature of Matlab, so
thank you in advance for any for any help!
> Alicia