Thread Subject: phase-calculation of disturbed data

Is there a way to get a reasonable phase-plot of a disturbed dataset?
I have already tried if I can get it right by changing the tolerance
of the function "unwrap", but it doesn't work. (The problem is that
"unwrap" only compares to the value before and doesn't take into
account a trend of the values around)


Thank you!

"Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
news:eebce67.-1@WebX.raydaftYaTP...
> Is there a way to get a reasonable phase-plot of a disturbed dataset?
> I have already tried if I can get it right by changing the tolerance
> of the function "unwrap", but it doesn't work. (The problem is that
> "unwrap" only compares to the value before and doesn't take into
> account a trend of the values around)
>
>
> Thank you!

Try the double-difference technique. I think it can work like this:
(1) compute the phase of each sample (which you have done)
(2) compute the phase difference from sample to sample
(3) call the UNWRAP function on that
(4) re-integrate the phases using CUMSUM

It's been a while, so my recollection of the implementation may be flawed,
but the concept is sound.
-Aj

In <eebce67.-1@WebX.raydaftYaTP> "Jochen Philipp" <jochen.philipp@gmx.de> writes:

>Is there a way to get a reasonable phase-plot of a disturbed dataset?
>I have already tried if I can get it right by changing the tolerance
>of the function "unwrap", but it doesn't work. (The problem is that
>"unwrap" only compares to the value before and doesn't take into
>account a trend of the values around)

One method would be to "smooth" the output from unwrap() by
selecting only the monotonic values from that output and
using interp1() with the 'pchip' option to fill in the
"errant" values.

>Thank you!

--
David B. Chorlian
Neurodynamics Lab SUNY/HSCB
chorlian@spot.cns.hscbklyn.edu
davidc@panix.com

AJ \no z\ Johnson wrote:
>
>
> "Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
> news:eebce67.-1@WebX.raydaftYaTP...
>> Is there a way to get a reasonable phase-plot of a disturbed
dataset?
>> I have already tried if I can get it right by changing the
tolerance
>> of the function "unwrap", but it doesn't work. (The problem is
that
>> "unwrap" only compares to the value before and doesn't take into
>> account a trend of the values around)
>>
>>
>> Thank you!
>
> Try the double-difference technique. I think it can work like this:
> (1) compute the phase of each sample (which you have done)
> (2) compute the phase difference from sample to sample
> (3) call the UNWRAP function on that
> (4) re-integrate the phases using CUMSUM
>
> It's been a while, so my recollection of the implementation may be
flawed,
> but the concept is sound.
> -Aj
>
>
>
The idea sounds good, but it doesn't work! That's the code I am using:


p=unwrap(unwrap(angle(G)));
phase=cumsum(p').*180./pi;


Anything wrong with that?


Regards, Jochen

>David B. Chorlian wrote:
>
> One method would be to "smooth" the output from unwrap() by
> selecting only the monotonic values from that output and
> using interp1() with the 'pchip' option to fill in the
> "errant" values.
>
>>Thank you!
>


Is there a way to find the monotonic values with a Matlab-Function?

"Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
news:eebce67.2@WebX.raydaftYaTP...
> AJ \no z\ Johnson wrote:
> >
> >
> > "Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
> > news:eebce67.-1@WebX.raydaftYaTP...
> >> Is there a way to get a reasonable phase-plot of a disturbed
> dataset?
> >> I have already tried if I can get it right by changing the
> tolerance
> >> of the function "unwrap", but it doesn't work. (The problem is
> that
> >> "unwrap" only compares to the value before and doesn't take into
> >> account a trend of the values around)
> >>
> >>
> >> Thank you!
> >
> > Try the double-difference technique. I think it can work like this:
> > (1) compute the phase of each sample (which you have done)
> > (2) compute the phase difference from sample to sample
> > (3) call the UNWRAP function on that
> > (4) re-integrate the phases using CUMSUM
> >
> > It's been a while, so my recollection of the implementation may be
> flawed,
> > but the concept is sound.
> > -Aj
> >
> >
> >
> The idea sounds good, but it doesn't work! That's the code I am using:
>
>
> p=unwrap(unwrap(angle(G)));
> phase=cumsum(p').*180./pi;
>
>
> Anything wrong with that?
>
>
> Regards, Jochen

Here's an example.
% Say you have a actual phase sequence that looks like this:
n=50;
p0=0.3*randn(1,n)+(0:n-1)*pi*0.95; subplot 311; plot(1:n,p0)
% But your measured phase looks like this:
p=mod(p0,2*pi);
% unwrap will have a very tough go of it.
subplot 312; plot(1:n,p,'b',1:n,unwrap(p),'r')
% So take the phase differences, unwrap, and reconstruct.
dp=diff(p);
subplot 313; plot(1:n,cumsum(unwrap([0,dp])))
% This may slope the wrong direction, because of freqeuncy
% wrapping at Nyquist frequency

May this will work in your application.
-Aj

"Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
news:<eebce67.-1@WebX.raydaftYaTP>...
> Is there a way to get a reasonable phase-plot of a disturbed dataset?
> I have already tried if I can get it right by changing the tolerance
> of the function "unwrap", but it doesn't work. (The problem is that
> "unwrap" only compares to the value before and doesn't take into
> account a trend of the values around)

See my post in the Aug '02 thread Re: Phase Unwrapping

Hope this helps.

Greg

In <eebce67.3@WebX.raydaftYaTP> "Jochen Philipp" <jochen.philipp@gmx.de> writes:

>>David B. Chorlian wrote:
>>
>> One method would be to "smooth" the output from unwrap() by
>> selecting only the monotonic values from that output and
>> using interp1() with the 'pchip' option to fill in the
>> "errant" values.
>>
>>>Thank you!
>>


>Is there a way to find the monotonic values with a Matlab-Function?

find(sign(diff(X)))
...

--
David B. Chorlian
Neurodynamics Lab SUNY/HSCB
chorlian@spot.cns.hscbklyn.edu
davidc@panix.com

"Jochen Philipp" <jochen.philipp@gmx.de> wrote in message news:<eebce67.-1@WebX.raydaftYaTP>...
> Is there a way to get a reasonable phase-plot of a disturbed dataset?
> I have already tried if I can get it right by changing the tolerance
> of the function "unwrap", but it doesn't work. (The problem is that
> "unwrap" only compares to the value before and doesn't take into
> account a trend of the values around)
>
>
> Thank you!

When we do phase averaging for block phase estimators in receivers we
usually convert the signal to real and imaginary. These values are
averaged, and then converted back to the phase domain.

HTH

Dick Walvis

I tried your suggested code on my data but it didn't work
sufficiently. Therefore I sent you an email with the code and my data
but you mail-address is not working. If you are still interested in
the problem, could you please mail me your correct address so I can
try it again?
Regards, Jochen
>
>

Jochen Philipp wrote:
>
>
> I tried your suggested code on my data but it didn't work
> sufficiently. Therefore I sent you an email with the code and my
data
> but you mail-address is not working. If you are still interested in
> the problem, could you please mail me your correct address so I can
> try it again?
> Regards, Jochen
>>
>>
>
uhhps! I forgot to mention which address I mean. It is:
aj.jozhnson@lmco.com

"Jochen Philipp" <jochen.philipp@gmx.de> wrote in message
news:eebce67.8@WebX.raydaftYaTP...
> I tried your suggested code on my data but it didn't work
> sufficiently. Therefore I sent you an email with the code and my data
> but you mail-address is not working. If you are still interested in
> the problem, could you please mail me your correct address so I can
> try it again?
> Regards, Jochen
> >
> >

It's de-spammed. Remove the "z".
-AJ "no Z" Johnson