In a previous thread I had a question about time domain aliasing and you
guys helped me out in understanding some. Mr. Lyons gave a really good
explanation. 
Previous thread:   http://www.dsprelated.com/showmessage/209160/1.php


I found the a thesis "Time Aliasing Methods of Spectrum Estimation" on some
of the topic. Good reading.... But questions just lead to more questions.
http://cdm15999.contentdm.oclc.org/cdm/singleitem/collection/ETD/id/16/rec/2


Ive also read papers on real time sliding DFTs of various sorts. My
question is for those implimentation gurus out there.

Would it be possible to impliment a 'decimated in frequency sliding dft'
efficiently in real time through the use of time domain aliasing?

If the operation involves time domain alising which is some form of overlap
and add, would that process have to be done off line, before the DFT is
done or is there a clever way to overlap the modulo N samples, drop off the
last one and add the newest one.

If it could be done in real time, it may be cool for some appllications
that have the need integer spaced samples of the DFT with more 'efficitve
frequency resolution'.

Any thoughts?	 

_____________________________		
Posted through www.DSPRelated.com