Quadratic interpolation of the signal frequency (DFT)

Hi all!
According to the Heisenberg Uncertainty principle, the frequency resolution
is limited in a finite time frame. However, if a sinusoid is the only
significant component in its vicinity, zero padding can be used to get a
better resolution of the DFT. This makes the spectral shape and place of the
sinusoid more clear and enables more accurate parameter estimation.

Does anybody know what type of window function should I use to get good
results?
Should I use to quadratic interpolate DFT value or log(DFT value)? I have
read that log(DFT value) gives better results, but I know why.

I would be grateful to have any explanation.

Cheers!
Andrzej

Hi!

Most interpolation methods are made for "rectangular" window.
And you can not use any other window anyway...
For zero padding, it does not matter...

Regards!
Janez.

"Andrzej Kaczor" <andrzejkaczor@wp.pl> wrote in message
news:bfe8rh$j35$1@news.onet.pl...
> Hi all!
> According to the Heisenberg Uncertainty principle, the frequency
resolution
> is limited in a finite time frame. However, if a sinusoid is the only
> significant component in its vicinity, zero padding can be used to get a
> better resolution of the DFT. This makes the spectral shape and place of
the
> sinusoid more clear and enables more accurate parameter estimation.
>
> Does anybody know what type of window function should I use to get good
> results?
> Should I use to quadratic interpolate DFT value or log(DFT value)? I have
> read that log(DFT value) gives better results, but I know why.
>
> I would be grateful to have any explanation.
>
> Cheers!
> Andrzej
>
>