Thank you for the help Steve. right now I am looking for how the scalable FFT is implemented in hardware. Because my interleaving factor is not a constant. It can change up to some maximum value. So having a hardwired approach maybe not be possible. Appreciate the help a lot Steve. Cheers.
Any FFT trick to do this?
Started by ●February 18, 2009
Reply by ●February 18, 20092009-02-18
Reply by ●February 18, 20092009-02-18
On Wed, 18 Feb 2009 01:35:04 -0600, "m26k9" <maduranga.liyanage@gmail.com> wrote:>I was looking at 'sliding DFT' approach.. looks like it has some hope.. >but I want to see if I can do it for steps larger than one. I.e. rather >than advancing the window by one sample, advance it by say N/2 samples.The sliding DFT makes use of recursion to reduce the computational load for the special case of windows advancing a sample at a time. If the window advances by more than one sample, the computations are no longer valid due to the missed steps in the recursion. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
Reply by ●February 18, 20092009-02-18
On Wed, 18 Feb 2009 12:54:08 -0700, Eric Jacobsen <eric.jacobsen@ieee.org> wrote:>The sliding DFT makes use of recursion to reduce the computational >load for the special case of windows advancing a sample at a time. If >the window advances by more than one sample, the computations are no >longer valid due to the missed steps in the recursion.A similar recursion can be formulated for "hops" greater than one sample. It's a straightforward derivation. Greg
Reply by ●February 18, 20092009-02-18
Greg Berchin <gberchin@comicast.net.invalid> wrote:><eric.jacobsen@ieee.org> wrote:>>The sliding DFT makes use of recursion to reduce the computational >>load for the special case of windows advancing a sample at a time. If >>the window advances by more than one sample, the computations are no >>longer valid due to the missed steps in the recursion.>A similar recursion can be formulated for "hops" greater than one >sample. It's a straightforward derivation.I do not think OP stated he needed a sliding DFT; just a DFT with low latency. Perhaps this could be clarified. Steve
Reply by ●February 19, 20092009-02-19
Thank you very much for all the replies. Actually when I thought more the sliding DFT or Geortzel algorithm do not work for me since I need to find a lot more than one bin of the FFt at a time. I do have a question on the scalable FFT. I will post a different thread on that since its not related to this. Thank you very much.