Has anyone seen echo patterns in the frequency (power) spectrum ?

Hi all

I m trying to fix this weird behavior that I see from a data obtained from
a terahertz spectroscopy analysis. The final data generated out of the
spectrometer software and i have perform some analysis on them. The first
step I need do is take fft of the data to able to see its features in
frequency domain but i observe a bizarre pattern. The power spectrum shows
echo type of patterns . This behavior is also consistent for various
different runs of the spectrometer. 

I have been doing a lot of search about this sort of problem but i m
clueless. Has anyone ever noticed this type of issue? If so , please point
me in right direction.

I would appreciate any input as i have been wracking my brains over this
for months now.

Thanks
Anu

anu22 wrote:

> Hi all
> 
> I m trying to fix this weird behavior that I see from a data obtained from
> a terahertz spectroscopy analysis.


Dear Academy of Sciences.
There is a strange underground knocking in the basement of my house. 
Because of that, I did not go down the basement for two years. This is 
unbearable. Let the scientists will explain me this knocking.

On 10/14/2011 1:04 PM, anu22 wrote:
> The power spectrum shows
> echo type of patterns .

Please tell us what that means, would you?

Fred

On Oct 14, 3:04&#4294967295;pm, "anu22" <anu.p.selvaraj@n_o_s_p_a_m.gmail.com>
wrote:

> The power spectrum shows
> echo type of patterns . This behavior is also consistent for various
> different runs of the spectrometer.

Are you trying to say that the spectra exhibit harmonic structures? If
so, then why are you surprised?

anu22 wrote:
> Hi all
>
> I m trying to fix this weird behavior that I see from a data obtained from
> a terahertz spectroscopy analysis.

Spectrogram of what? Of cosmic background? Of a sample whose 
structure &/or composition to be determined? I suspect the 
later.


> The final data generated out of the
> spectrometer software and i have perform some analysis on them. The first
> step I need do is take fft of the data to able to see its features in
> frequency domain but i observe a bizarre pattern. The power spectrum shows
> echo type of patterns . This behavior is also consistent for various
> different runs of the spectrometer.

If it is consistent you may be unexpectedly measuring a 
physical parameter of your setup. What is characteristic 
time delay of the "echo"? Considering the mediums 
propagation velocity, does the corresponding distance match 
something in your setup.

I remember many years ago one of my tasks as an Instrument 
Technician was to run NMR spectra for those outside the 
department. This one biochemist kept giving the same sample 
size ass he gave the mass spectrometer team. He wondered why 
I kept giving him very nice spectra of the hydrogen impurity 
of his deuterated solvent ;/




>
> I have been doing a lot of search about this sort of problem but i m
> clueless. Has anyone ever noticed this type of issue? If so , please point
> me in right direction.
>
> I would appreciate any input as i have been wracking my brains over this
> for months now.
>
> Thanks
> Anu
>
>

Hi

Sorry about getting back late. 
its not harmonics structure. And its possible maybe i m measuring a
convoluted signal which has my actual signal and some other signal (maybe
from the measuring probe or something).I would liek to show the spectra
plot but this is my first time posting and i m not sure how to do that?

if anyone can tell me that , then i can show my plot versus teh expected
plot .


>anu22 wrote:
>> Hi all
>>
>> I m trying to fix this weird behavior that I see from a data obtained
from
>> a terahertz spectroscopy analysis.
>
>Spectrogram of what? Of cosmic background? Of a sample whose 
>structure &/or composition to be determined? I suspect the 
>later.
>
>
>> The final data generated out of the
>> spectrometer software and i have perform some analysis on them. The
first
>> step I need do is take fft of the data to able to see its features in
>> frequency domain but i observe a bizarre pattern. The power spectrum
shows
>> echo type of patterns . This behavior is also consistent for various
>> different runs of the spectrometer.
>
>If it is consistent you may be unexpectedly measuring a 
>physical parameter of your setup. What is characteristic 
>time delay of the "echo"? Considering the mediums 
>propagation velocity, does the corresponding distance match 
>something in your setup.
>
>I remember many years ago one of my tasks as an Instrument 
>Technician was to run NMR spectra for those outside the 
>department. This one biochemist kept giving the same sample 
>size ass he gave the mass spectrometer team. He wondered why 
>I kept giving him very nice spectra of the hydrogen impurity 
>of his deuterated solvent ;/
>
>
>
>
>>
>> I have been doing a lot of search about this sort of problem but i m
>> clueless. Has anyone ever noticed this type of issue? If so , please
point
>> me in right direction.
>>
>> I would appreciate any input as i have been wracking my brains over
this
>> for months now.
>>
>> Thanks
>> Anu
>>
>>
>

On 10/18/2011 10:56 AM, anu22 wrote:
> Hi
>
> Sorry about getting back late.
> its not harmonics structure. And its possible maybe i m measuring a
> convoluted signal which has my actual signal and some other signal (maybe
> from the measuring probe or something).I would liek to show the spectra
> plot but this is my first time posting and i m not sure how to do that?
>
> if anyone can tell me that , then i can show my plot versus teh expected
> plot .

Put your plot on an accessible web site and post a link to it.

Jerry
-- 
Engineering is the art of making what you want from things you can get.

thanks jerry.

i have the links for the spectrum so that u can see that i talking abt .
they are not supposed to be identical but look generally similar.
this is the one for the ideal spectrum
http://www.scribd.com/doc/69446056/freqdomainsnrideal

where as mine looks like this :
http://www.scribd.com/doc/69445630/freqdomainsnrdb

regards
Anu

>On 10/18/2011 10:56 AM, anu22 wrote:
>> Hi
>>
>> Sorry about getting back late.
>> its not harmonics structure. And its possible maybe i m measuring a
>> convoluted signal which has my actual signal and some other signal
(maybe
>> from the measuring probe or something).I would liek to show the spectra
>> plot but this is my first time posting and i m not sure how to do that?
>>
>> if anyone can tell me that , then i can show my plot versus teh
expected
>> plot .
>
>Put your plot on an accessible web site and post a link to it.
>
>Jerry
>-- 
>Engineering is the art of making what you want from things you can get.
>

On Wed, 19 Oct 2011 09:37:29 -0500, "anu22"
<anu.p.selvaraj@n_o_s_p_a_m.gmail.com> wrote:

>i have the links for the spectrum so that u can see that i talking abt .
>they are not supposed to be identical but look generally similar.
>this is the one for the ideal spectrum
>http://www.scribd.com/doc/69446056/freqdomainsnrideal
>
>where as mine looks like this :
>http://www.scribd.com/doc/69445630/freqdomainsnrdb

First and foremost: you need to put them both on the same scale, especially in
the horizontal.

Second: these are not spectra, they are ratios of spectra -- ratio of signal
spectrum to noise spectrum. Your measurements are evidently very noisy, and at
frequencies where you have larger-than-expected noise components, your SNR is
reduced (and vice-versa).

Taking all of this into account, the measured curve does not look all that
unreasonable to me. Can you eliminate some noise sources and repeat the
measurements, to be certain?

Greg

the horizontal scale can be made till 7 thz but i expanded to 10 so that
everyone can see the entire replication. 
Also in the second spectrum the long dips that you see are due the features
of the sample and it has to be there. I think i gave wrong examples:

please disregard the prev links and use these as an example of what i m
describing:
http://www.scribd.com/doc/69458803/Freqdomain-Ideal

the same data when i perform my processing looks like this:
http://www.scribd.com/doc/69458941/freqdomain

Also the second plot is till 20thz cos i want to show the how the features
get repetitive .


>On Wed, 19 Oct 2011 09:37:29 -0500, "anu22"
><anu.p.selvaraj@n_o_s_p_a_m.gmail.com> wrote:
>
>>i have the links for the spectrum so that u can see that i talking abt .
>>they are not supposed to be identical but look generally similar.
>>this is the one for the ideal spectrum
>>http://www.scribd.com/doc/69446056/freqdomainsnrideal
>>
>>where as mine looks like this :
>>http://www.scribd.com/doc/69445630/freqdomainsnrdb
>
>First and foremost: you need to put them both on the same scale,
especially in
>the horizontal.
>
>Second: these are not spectra, they are ratios of spectra -- ratio of
signal
>spectrum to noise spectrum. Your measurements are evidently very noisy,
and at
>frequencies where you have larger-than-expected noise components, your SNR
is
>reduced (and vice-versa).
>
>Taking all of this into account, the measured curve does not look all
that
>unreasonable to me. Can you eliminate some noise sources and repeat the
>measurements, to be certain?
>
>Greg
>