Forums

Shifting data before FFT

Started by Andy April 28, 2009
I'm doing some DSP for radar applications.  My background in DSP is
limited, so I'm trying to get up to speed.

Performing a circular rotation to move the beginning of data to the
center of the dataset before a FFT and then rotating back after seems
to be routine (almost taken for granted).  Can anyone explain the
benefit of this when compared to just a straight FFT of the data?


Thank you

Andy
On Apr 28, 9:20 am, Andy <andya...@gmail.com> wrote:
> I'm doing some DSP for radar applications. My background in DSP is > limited, so I'm trying to get up to speed. > > Performing a circular rotation to move the beginning of data to the > center of the dataset before a FFT and then rotating back after seems > to be routine (almost taken for granted). Can anyone explain the > benefit of this when compared to just a straight FFT of the data? > > Thank you > > Andy
Writing in terms of time series at the input: Do you prefer to view t=0 at the start of the array or in the middle? Do you prefer to get DC out at the start of the array or in the middle? How does your fft function behave? Rotate as necessary. Dale B. Dalrymple
On 28 Apr, 18:20, Andy <andya...@gmail.com> wrote:
> I'm doing some DSP for radar applications. &#2013266080;My background in DSP is > limited, so I'm trying to get up to speed. > > Performing a circular rotation to move the beginning of data to the > center of the dataset before a FFT and then rotating back after seems > to be routine (almost taken for granted). &#2013266080;Can anyone explain the > benefit of this when compared to just a straight FFT of the data?
As others already indicated, a lot of this is a matter of personal taste or established convension in particular fields. However, in the case of radar and sonar, it might make a certain sense to organize the spectrum with DC near the middle. Human users might find it easier to spot and interpret Doppler shifts in such plots. Rune
On Apr 30, 10:21&#2013266080;am, Rune Allnor <all...@tele.ntnu.no> wrote:
> On 28 Apr, 18:20, Andy <andya...@gmail.com> wrote: > > > I'm doing some DSP for radar applications. &#2013266080;My background in DSP is > > limited, so I'm trying to get up to speed. > > > Performing a circular rotation to move the beginning of data to the > > center of the dataset before a FFT and then rotating back after seems > > to be routine (almost taken for granted). &#2013266080;Can anyone explain the > > benefit of this when compared to just a straight FFT of the data? > > As others already indicated, a lot of this is a matter > of personal taste or established convension in particular > fields. > > However, in the case of radar and sonar, it might > make a certain sense to organize the spectrum with > DC near the middle. Human users might find it easier > to spot and interpret Doppler shifts in such plots. > > Rune
Thank you Dale and Rune for the suggestions. I think that canceling the Doppler shifts may just be the key here.
On Apr 30, 7:21 am, Rune Allnor <all...@tele.ntnu.no> wrote:
> On 28 Apr, 18:20, Andy <andya...@gmail.com> wrote: > > > I'm doing some DSP for radar applications. My background in DSP is > > limited, so I'm trying to get up to speed. > > > Performing a circular rotation to move the beginning of data to the > > center of the dataset before a FFT and then rotating back after seems > > to be routine (almost taken for granted). Can anyone explain the > > benefit of this when compared to just a straight FFT of the data? > > As others already indicated, a lot of this is a matter > of personal taste or established convension in particular > fields. >
.> However, in the case of radar and sonar, it might .> make a certain sense to organize the spectrum with .> DC near the middle. Human users might find it easier .> to spot and interpret Doppler shifts in such plots. .> .> Rune That's a good example. but not because it is radar and sonar but because it is a case where the processed band has been complex demodulated to baseband and contains frequencies both higher and lower than what is now represented by DC. That is almost always true in radar and active sonar. In passive sonar, audio and music the signal is often real with no different information on one side of DC than the other. Even in the bandpass cases, where you want to put DC may depend on whether or not the sidebands are symmetric. Dale B. Dalrymple