Zero padded FFT

I have a query on zero-padded FFT's.For calculating FFT of a signal
with  sampling rate 256 HZ  what is the maximum zero padding that can
be done?And what does zeropadding actually do to the signal.  I have
observed frequency shift with zero padding.why so?

Zero padding actually more or less interpolates the frequency domain
points you would have without zero padding. Frequency shifts are
impossible in this respect.
I don't really know what you mean by 'maximum zero padding'...

Ciao,
Dominiek


ssahuiit@rediffmail.com wrote in message news:<2ce2ab3c.0307040100.4b2dddc2@posting.google.com>...
> I have a query on zero-padded FFT's.For calculating FFT of a signal
> with  sampling rate 256 HZ  what is the maximum zero padding that can
> be done?And what does zeropadding actually do to the signal.  I have
> observed frequency shift with zero padding.why so?

ssahuiit@rediffmail.com wrote in message news:<2ce2ab3c.0307040100.4b2dddc2@posting.google.com>...
> I have a query on zero-padded FFT's.For calculating FFT of a signal
> with  sampling rate 256 HZ  what is the maximum zero padding that can
> be done?

Zero padding appends artificial data points of value 0 to the 
finite data series. The amount of artificial data points that 
can be appended would probably be limited by system memory. 

> And what does zeropadding actually do to the signal.  

It interpolates the spectrum.

> I have
> observed frequency shift with zero padding.why so?

The values of the DFT coefficient are constrained on the spectrum 
lines that also are found in the DFT of the non-zero padded sequence. 
However, the gradient is not, so one might see situations where 
the interpolated spectrum has a positive gradient dX/df on one "true"
line and a negative gradient on the nex, which means there would be a 
maximum in between. It's can be regarded as an artifact of the 
interpolation.

Rune