Windowing question

Hi, I noticed this phenomena when using a Hanning window before using
FFT of 4096 size.
If the data that I put into the FFT is up to about 2048 samples,
Hanning window reduces the side-lobes as expected. But if I use more
data on the FFT such as 4096 samples, the Hanning window actualy add
more side-lobes than Rectangular. Is this the way it supposed to work?
I used Sound-Forge spectrum analisys for these tests. maybe using
rectangular window the side-lobes are so nerrow that this spectrum
analiser don't show them? I did notice that it ignored very nerrow
peaks.
Like I said the actual FFT size used on these test was 4096. 

Thanks,

Oded Streigold.

Oded wrote:

> Hi, I noticed this phenomena when using a Hanning window before using
> FFT of 4096 size.

Just a nit, but the correct term is "Hann" or "von Hann", not "Hanning", 
after Austrian meteorologist Julius von Hann.

OUP

"Oded" <o@sadglad.com> wrote in message
news:7c5c69e0.0308141013.3e7c3167@posting.google.com...
> Hi, I noticed this phenomena when using a Hanning window before using
> FFT of 4096 size.
> If the data that I put into the FFT is up to about 2048 samples,
> Hanning window reduces the side-lobes as expected. But if I use more
> data on the FFT such as 4096 samples, the Hanning window actualy add
> more side-lobes than Rectangular. Is this the way it supposed to work?
> I used Sound-Forge spectrum analisys for these tests. maybe using
> rectangular window the side-lobes are so nerrow that this spectrum
> analiser don't show them? I did notice that it ignored very nerrow
> peaks.
> Like I said the actual FFT size used on these test was 4096.

If the FFT size was 4096, what was the window length?  If the window
remained short, then this might be the result.

Also, if the input data naturally has energy at certain frequencies and now
you're resolving them ..... that could account for higher apparent
"sidelobes".

Fred