low freqeuncy peaks in FFT

Together with a friend, I am doing research on acoustics in neolithic sites and in the recording of silence in such neolithic passage we see a low frequency component (6, 11, 17 Hz), sometimes also 8 Hz, etc depending on the passage or(?) situation.. (and this 8 Hz could map on the resonance frequency of a passage of around 11 m).
These low freq. peaks do not only happen in neolithic buildings, but also in my modern drawing room... The peak density depends on the sampling rate and bitdepth:
. 96kHz seems to have half the density of 48kHz.
. 16bits bit depth seems to have a 2/3 density compared to 24bits bit depth.
<by the way I uses an Zoom H1 recorder and Audacity 2.0.5 for analysis>

I have the idea that these low frequency peaks are side lobes of a sinc function, but not sure... So why do they emerge in the spectrum (using FFT)?

Do you have a clue? Any hints/reference/links are very welcome. 
I can provide an audio file and a picture if wanted (although how to I add a picture to my post?).
Thanks in advance for your help!

All the best,


Victor

Sometimes I am wondering if this could be due to a difference in the sampling rate in my recorder, which can of course be out with a few Hz. And this error is seen as peaks in the FFT?
But I am not sure about this...


On Friday, 20 November 2015 13:04:16 UTC, Victor Reijs  wrote:
> Together with a friend, I am doing research on acoustics in neolithic sites and in the recording of silence in such neolithic passage we see a low frequency component (6, 11, 17 Hz), sometimes also 8 Hz, etc depending on the passage or(?) situation.. (and this 8 Hz could map on the resonance frequency of a passage of around 11 m).
> These low freq. peaks do not only happen in neolithic buildings, but also in my modern drawing room... The peak density depends on the sampling rate and bitdepth:
> . 96kHz seems to have half the density of 48kHz.
> . 16bits bit depth seems to have a 2/3 density compared to 24bits bit depth.
> <by the way I uses an Zoom H1 recorder and Audacity 2.0.5 for analysis>

Have you considered the source of the noise could be electronic within your recording equipment.

Disconnect the mic and record true silence and see what you get.

Mark

On Friday, 20 November 2015 14:26:08 UTC, mako...@yahoo.com  wrote:
> Have you considered the source of the noise could be electronic within your recording equipment.
> 
> Disconnect the mic and record true silence and see what you get.
> 
> Mark

Thanks Mark, I indeed did that now and I see similar peaks incl. the sinc- function (it has peaks at n*fo, with f0 around 3 Hz) (so no mic connected to the input but kept recording level same). So this is coming from the internal electronics of the recorder. Thanks for pointing this out!

Is there any idea how to remove that (except buying other equipment)?
I am wondering how this influences the low frequency analysis (say from 10 Hz, which is mentioned as the freq range of the involved mic of the recorder)? So perhaps I need to do two recordings with different bitdepth, and then the difference provides a better idea on the actual sound...

Any ideas how to improve the analysis of an FFT here? Thanks in advance.

All the best,


Victor

On Fri, 20 Nov 2015 07:05:13 -0800 (PST), Victor Reijs
<victor.reijs@gmail.com> wrote:

>On Friday, 20 November 2015 14:26:08 UTC, mako...@yahoo.com  wrote:
>> Have you considered the source of the noise could be electronic within yo=
>ur recording equipment.
>>=20
>> Disconnect the mic and record true silence and see what you get.
>>=20
>> Mark
>
>Thanks Mark, I indeed did that now and I see similar peaks incl. the sinc- =
>function (it has peaks at n*fo, with f0 around 3 Hz) (so no mic connected t=
>o the input but kept recording level same). So this is coming from the inte=
>rnal electronics of the recorder. Thanks for pointing this out!
>
>Is there any idea how to remove that (except buying other equipment)?
>I am wondering how this influences the low frequency analysis (say from 10 =
>Hz, which is mentioned as the freq range of the involved mic of the recorde=
>r)? So perhaps I need to do two recordings with different bitdepth, and the=
>n the difference provides a better idea on the actual sound...
>
>Any ideas how to improve the analysis of an FFT here? Thanks in advance.
>
>All the best,

Could you maybe post a graph of what you're seeing?    If there is a
DC offset, even a small one, it could be sidelobes from that.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On Friday, 20 November 2015 16:25:38 UTC, Eric Jacobsen  wrote:
> Could you maybe post a graph of what you're seeing?    If there is a
> DC offset, even a small one, it could be sidelobes from that.

How do I send a picture on this group? 
I used in Audacity Effects -> Normalize... -> 'Remove DC offset' before the analusis.

I understand in MATLAB one has the function 'detrend' but that is not available in Audacity...

All the best,

Victor

On Fri, 20 Nov 2015 08:57:04 -0800 (PST), Victor Reijs
<victor.reijs@gmail.com> wrote:

>On Friday, 20 November 2015 16:25:38 UTC, Eric Jacobsen  wrote:
>> Could you maybe post a graph of what you're seeing?    If there is a
>> DC offset, even a small one, it could be sidelobes from that.
>
>How do I send a picture on this group? 
>I used in Audacity Effects -> Normalize... -> 'Remove DC offset' before the analusis.
>
>I understand in MATLAB one has the function 'detrend' but that is not available in Audacity...
>
>All the best,
>
>Victor

You can post a file to a cloud server (dropbox, google drive, etc.,
etc.), or any site with public read acces,  make it public and post a
link here.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On Friday, 20 November 2015 18:57:13 UTC, Eric Jacobsen  wrote:
> On Fri, 20 Nov 2015 08:57:04 -0800 (PST), Victor Reijs
> >How do I send a picture on this group? 

> You can post a file to a cloud server (dropbox, google drive, etc.,
> etc.), or any site with public read acces,  make it public and post a
> link here.

I hope this is visible:
https://goo.gl/photos/3GP3hcoyxuSUvKqTA

On Fri, 20 Nov 2015 11:43:15 -0800 (PST), Victor Reijs
<victor.reijs@gmail.com> wrote:

>On Friday, 20 November 2015 18:57:13 UTC, Eric Jacobsen  wrote:
>> On Fri, 20 Nov 2015 08:57:04 -0800 (PST), Victor Reijs
>> >How do I send a picture on this group? 
>
>> You can post a file to a cloud server (dropbox, google drive, etc.,
>> etc.), or any site with public read acces,  make it public and post a
>> link here.
>
>I hope this is visible:
>https://goo.gl/photos/3GP3hcoyxuSUvKqTA

Yes, that's visible.

Those don't looke like sinc sidelobes.   They more closely resemble
harmonics or are perhaps some sort of processing artifact.   If you
can't completely eliminate acoustic stimulus, which I'd guess is hard
to do, could that be harmonic resonance of the environment?   I'm
guessing there.

Is the vertical scale dBFS?  What is the reference for the dB scale?


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On Friday, November 20, 2015 at 2:43:24 PM UTC-5, Victor Reijs wrote:
> On Friday, 20 November 2015 18:57:13 UTC, Eric Jacobsen  wrote:
> > On Fri, 20 Nov 2015 08:57:04 -0800 (PST), Victor Reijs
> > >How do I send a picture on this group? 
> 
> > You can post a file to a cloud server (dropbox, google drive, etc.,
> > etc.), or any site with public read acces,  make it public and post a
> > link here.
> 
> I hope this is visible:
> https://goo.gl/photos/3GP3hcoyxuSUvKqTA

yes it is.

Is the y axis dB below full scale?

If so, those are pretty low levels.

What kind of recorder is it.

The noise may be unwanted crosstalk from the digtal circuits into the 
analog.  If they are below the frequency range of your interest, high pass them away.

If the acoustical signals you want to study are very low level, then you may benefit from an external preamp to increase the electrical signal level before it enters your recorder.   That will relativly reduce the level of this unwanted noise.


Mark