Re: [LAD] Listing lowest and highest frequencies in a track?

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: <linux-audio-dev@...>
Date: Friday, August 31, 2012 - 6:48 pm

--e89a8f647b1df2079e04c8943a13
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On Fri, Aug 31, 2012 at 6:40 PM, Paul Davis wro=
te:

> On Fri, Aug 31, 2012 at 11:26 AM, Chris Bannister <

ote:
t
If we really start to look at the details of this question, the dB issue is
the least of concerns, but let's look at that first:

Paul probably meant sample value 1.0 to be 0 dBFS. That is a clear and good
definition for sample values, but powers (RMS), are not that simple: Some
like to keep things a simple and just treat the signal and power levels
equally, giving a full scale square wave the power of 0 dB. However, this
leads to the fact that a sine wave (and thus also an isolated frequency)
can have a power of -3 dB at the maximum. Some like to make things a bit
more complicated, and define power dB relative to the power of a full scale
sine wave.

However, the biggest problem in the question is that it doesn't consider
the time-frequency uncertainty, and the fundamental nature of time limited
signals (a time limited signal can't be band limited).

You can not measure frequencies whose period is shorter than the
measurement data. That means that you can't measure the power at 1Hz with a
resolution better than one second. This means that the "where they are"
part of the question is not well defined.

If you take one sample from the signal, and analyse that, you'll just have
an impulse. And an impulse has equal power at frequencies from 0 to
nyquist. The problem we see here will manifest itself with any time limited
signal, you will have some "leak" which will spread all across the
spectrum. This means the "lowest and highest" part of the question doesn't
make sense: it will always be from zero (or the lowest bin) to nyquist.

What you can do, is use a tool like Sonic Visualiser to look at the
spectrogram of the piece (with long overlapping analysis windows). Playing
around with the analysis settings should also teach you about the
time-frequency uncertainty I discussed above, in a rather interactive way.
It also includes nice stuff like peak frequency plotting, certainly worth a
look at.

-Sakari-

--e89a8f647b1df2079e04c8943a13
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On Fri, Aug 31, 2012 at 6:40 PM, Paul Davis <paul@linuxaudiosys=
tems.com
> wrote:
On Fri, Aug 31, 2012 at 11:26 =
AM, Chris Bannister <cbannister@slingshot.co.nz> wr=
ote:

On Mon, Aug 27, 2012 at 10:54:28PM -0600, Bearcat M. =C5=9E=C3=A1ndor =
wrote:

t
br>

Do you mean dBm? dB is a ratio.dbFS p=
robably, since its digital (sample value =3D=3D 0 =3D> 0 dbFS)=

If we really start to look at the details o=
f this question, the dB issue is the least of concerns, but let's look =
at that first:Paul probably meant sample value 1.=
0 to be 0 dBFS. That is a clear and good definition for sample values, but =
powers (RMS), are not that simple:=C2=A0Some like to keep things a simple a=
nd just treat the signal and power levels equally, giving a full scale squa=
re wave the power of 0 dB. However, this leads to the fact that a=C2=A0sine=
wave (and thus also an isolated frequency) can have a power of -3 dB at th=
e maximum. Some like to make things a bit more complicated, and define powe=
r dB relative to the power of a full scale sine wave.
However, the biggest problem in the question is that it=
doesn't consider the time-frequency=C2=A0uncertainty, and the fundamen=
tal nature of time limited signals (a time limited signal can't be band=
limited).=C2=A0
You can not measure frequencies whose period is shorter=
than the measurement data. That means that you can't measure the power=
at 1Hz with a resolution better than one second. This means that the &quot=
;where they are" part of the question is not well defined.
If you take one sample from the signal, and analyse tha=
t, you'll just have an impulse. And an impulse has equal power at frequ=
encies from 0 to nyquist. The problem we see here will manifest itself with=
any time limited signal, you will have some "leak" which will sp=
read all across the spectrum. This means the "lowest and highest"=
part of the question doesn't make sense: it will always be from zero (=
or the lowest bin) to nyquist.
What you can do, is use a tool like Sonic Visualiser to=
look at the spectrogram of the piece (with long overlapping analysis windo=
ws). Playing around with the analysis settings should also teach you about =
the time-frequency uncertainty I discussed above, in a rather interactive w=
ay. It also includes nice stuff like peak frequency plotting, certainly wor=
th a look at.
-Sakari-

--e89a8f647b1df2079e04c8943a13--

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [LAD] Listing lowest and highest frequencies in a track?, Chris Bannister, (Fri Aug 31, 3:34 pm)
Re: [LAD] Listing lowest and highest frequencies in a track?, Sakari Bergen, (Fri Aug 31, 6:48 pm)
Re: [LAD] Listing lowest and highest frequencies in a track?, Fons Adriaensen, (Fri Aug 31, 8:11 pm)
Re: [LAD] Listing lowest and highest frequencies in a track?, Jens M Andreasen, (Wed Sep 5, 10:57 am)
Re: [LAD] Listing lowest and highest frequencies in a track?, Harry van Haaren, (Tue Aug 28, 11:52 am)