Hi, I have a question regarding the deconvolution of a signal from the window. we had non-equispaced vlbi data, to do the analysis we first windowed the data in the time domain using parseval window then applied SVD and DFT in order to get the power spectrum in the frequency domain. But using the parseval window reduced our resolution and broadened our peak. Since we are interested in Q-factor value of the peak we want to deconvolve the signal from the window. Is it possible to reverse the effect of parseval window completely from the data and get the original signal in frequency domain. Thank you. Midhat.

# Deconvolution

Started by ●August 7, 2008

Reply by ●August 7, 20082008-08-07

On 7 Aug., 18:57, "Midhat" <midhatzub...@gmail.com> wrote:> Hi, > I have a question regarding the deconvolution of a signal from the window. > we had non-equispaced vlbi data, to do the analysis we first windowed the > data in the time domain using parseval window then applied SVD and DFT in > order to get the power spectrum in the frequency domain. But using the > parseval window reduced our resolution and broadened our peak. Since we are > interested in Q-factor value of the peak we want to deconvolve the signal > from the window. Is it possible to reverse the effect of parseval window > completely from the data and get the original signal in frequency domain. > Thank you. > Midhat.ifft -> divide by window -> fft. If your window has zeros in it (perhapst at the ends) then you irrevocably destroyed the time domain data. Regards, Andor

Reply by ●August 7, 20082008-08-07

Midhat wrote:> I have a question regarding the deconvolution of a signal from the window. > we had non-equispaced vlbi data, to do the analysis we first windowed the > data in the time domain using parseval window then applied SVD and DFT in > order to get the power spectrum in the frequency domain.I have Jansson's "Deconvolution of Images and Spectra" http://www.amazon.com/Deconvolution-Images-Spectra-Peter-Jansson/dp/0123802229 which mostly covers non-linear deconvolution. Consider that emission can't go negative, and absorption spectra can't go greater than one or negative, but linear deconvolution in the presence of noise can easily violate those restrictions. There may be newer books, but I would still recommend looking at this one. -- glen

Reply by ●August 7, 20082008-08-07

On Aug 7, 9:57 am, "Midhat" <midhatzub...@gmail.com> wrote:> Hi, > I have a question regarding the deconvolution of a signal from the window. > we had non-equispaced vlbi data, to do the analysis we first windowed the > data in the time domain using parseval window then applied SVD and DFT in > order to get the power spectrum in the frequency domain. But using the > parseval window reduced our resolution and broadened our peak. Since we are > interested in Q-factor value of the peak we want to deconvolve the signal > from the window. Is it possible to reverse the effect of parseval window > completely from the data and get the original signal in frequency domain. > Thank you. > Midhat.So what are you using (mathematical function) as a "parseval window"? Dale B. Dalrymple