time difference between two acoustic sine waves

i have two physically separated acoustic sensors to pick up acoustic
signals..when they pick up a signal , due to differences in distance
covered , there is a phase difference for the corresponding time delay..As
of now I am using correlation in time domain to calculate the time delay
..But i want to explore the method of Zero crossing for the same or fine
tune the correlation method..Can someone please guide me on this...thanks
and regards

dks

On Nov 26, 10:37 am, "rashdk" <sing...@rediffmail.com> wrote:
> i have two physically separated acoustic sensors to pick up acoustic
> signals..when they pick up a signal , due to differences in distance
> covered , there is a phase difference for the corresponding time delay..As
> of now I am using correlation in time domain to calculate the time delay
> ..But i want to explore the method of Zero crossing for the same or fine
> tune the correlation method..Can someone please guide me on this...thanks
> and regards
>
> dks

The only place that I know of where a clipped correlator works better
is when your noise is highly impulsive, like under deep pack ice.

If you really want to look at clipped correlators, you should look at
old JASA stuff, late 50's to mid 60's.

On 26 Nov, 16:37, "rashdk" <sing...@rediffmail.com> wrote:
> i have two physically separated acoustic sensors to pick up acoustic
> signals..when they pick up a signal , due to differences in distance
> covered , there is a phase difference for the corresponding time delay..As
> of now I am using correlation in time domain to calculate the time delay
> ..But i want to explore the method of Zero crossing for the same or fine
> tune the correlation method..Can someone please guide me on this...thanks
> and regards

In time domain you can only locate a correlation peak or a
zero crossing to within one sample period, T. If you want
better location in time you might want to try to estimate
the slope of the phase of the cross spectrum.

You would need a broad-band signal for that, though.

Rune

On Nov 26, 7:37 am, "rashdk" <sing...@rediffmail.com> wrote:
> i have two physically separated acoustic sensors to pick up acoustic
> signals..when they pick up a signal , due to differences in distance
> covered , there is a phase difference for the corresponding time delay..As
> of now I am using correlation in time domain to calculate the time delay
> ..But i want to explore the method of Zero crossing for the same or fine
> tune the correlation method..Can someone please guide me on this...thanks
> and regards
>
> dks

dks

Once you  have properly sampled your signals, you can use those
samples to reconstruct the signals at other points in time. If you
resample your signals at smaller intervals, picking the sample time
when the signal is nearest to zero will give the opportunity to pick
at smaller intervals. This can reduce the error in estimation of zero
crossing time due to the time quantization of your original samples.
It will not reduce the error in estimation of zero crossing time due
to limited signal to noise ratio of your original samples.

Time samples at smaller intervals can also be used to produce
correlation estimates at smaller intervals.  This can reduce the error
in estimation of correlation peak time due to the time quantization of
your original samples. It will not reduce the error in estimation of
correlation peak time due to limited signal to noise ratio of your
original samples.

There are also methods that can be used to estimate the location of
zero crossings (or correlation peaks) from samples near the zero
crossing (or peak). These usually require additional knowledge of the
signal. For example,if you know the signal is a pure large sine wave,
samples near the zero crossing may be approximately  linear. For
correlation peaks a parabolic interpolation might be used.

Transform domain techniques exist that can provide efficient
implementations of the resampling. Ask google.

Good luck,

Dale B. Dalrymple
http://dbdimages.com
http://stores.lulu.com/dbd

On 27 Nov, 07:03, dbd <d...@ieee.org> wrote:

> Transform domain techniques exist that can provide efficient
> implementations of the resampling.

You wouldn't happen to be thinking about something like this:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
T1 = 1;
P = 8;
T2 = T1/P;

N = 8;
t = 0:(N-1)*T1;
f=1/8;                               % with this f the results look
nice
%f = 1/10;                         % with this f the results look...?

s = sin(2*pi*f*t);

S = fft(s);
M = P*N-floor(N/2)-1;
Si = [S(1:floor(N/2)+1),zeros(1,M)];
si = 2*real(ifft(Si));
ti = (0:N*P-1)*T2;

clf
s1=stem(t,s); set(s1,'color','blue')
hold on
s2=stem(ti,si);set(s2,'color','red')

%%%%%%%%%%%%%%%%%%%%%%%%%%

Rune

On Nov 27, 12:21 am, Rune Allnor <all...@tele.ntnu.no> wrote:
> On 27 Nov, 07:03, dbd <d...@ieee.org> wrote:
>
> > Transform domain techniques exist that can provide efficient
> > implementations of the resampling.
>
> You wouldn't happen to be thinking about something like this:
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> T1 = 1;
> P = 8;
> T2 = T1/P;
>
> N = 8;
> t = 0:(N-1)*T1;
> f=1/8;                               % with this f the results look
> nice
> %f = 1/10;                         % with this f the results look...?
>
> s = sin(2*pi*f*t);
>
> S = fft(s);
> M = P*N-floor(N/2)-1;
> Si = [S(1:floor(N/2)+1),zeros(1,M)];
> si = 2*real(ifft(Si));
> ti = (0:N*P-1)*T2;
>
> clf
> s1=stem(t,s); set(s1,'color','blue')
> hold on
> s2=stem(ti,si);set(s2,'color','red')
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%
>
> Rune

Rune

I would never call you google.

Or do dks' homework.

Dale B. Dalrymple

While it's topic may seem completely different from what you are asking for
you should have a look at

tda.jpl.nasa.gov/progress_report/42-121/121G.pdf

and specially at the simple math presented in the appendices A, B and C.

Regards
Ulrich Bangert

"rashdk" <singhdk@rediffmail.com> schrieb im Newsbeitrag
news:9uqdncgm9uPYddfanZ2dnUVZ_saknZ2d@giganews.com...
> i have two physically separated acoustic sensors to pick up acoustic
> signals..when they pick up a signal , due to differences in distance
> covered , there is a phase difference for the corresponding time delay..As
> of now I am using correlation in time domain to calculate the time delay
> .But i want to explore the method of Zero crossing for the same or fine
> tune the correlation method..Can someone please guide me on this...thanks
> and regards
>
> dks

I have some code for the correlation method here:

http://www.dsprelated.com/showarticle/26.php

Not sure if it qualifies as "fine-tuned", but it reaches accuracy of
10e-10 sample time if the signal is noise free.

-mn

On 28 Nov, 12:02, "mnentwig" <mnent...@elisanet.fi> wrote:
> I have some code for the correlation method here:
>
> http://www.dsprelated.com/showarticle/26.php
>
> Not sure if it qualifies as "fine-tuned", but it reaches accuracy of
> 10e-10 sample time if the signal is noise free.

Very intersting page. You probably don't know it (or you
would have referenced it already) but this stuff is treated in

Bendat & Piersol: "Random Data", Wiley, 20090,

which is a classic in data analysis.

Rune

>You probably don't know it (or you
>would have referenced it already) but this stuff is treated in

Thanks for the reference, and no, I didn't know it. I'll add it to the
blog.

Cheers

Markus