ee05s030 wrote on Jul 26th, 2006, 10:38pm:when we find the fft of some data, what does it actually mean? how can we get the the energy density spectrum ot power density spectrum from the fft output? what are the scaling factors needed?
I assume you are talking about noise. The noise spectral density has units of V^2/Hz. At each measurement frequency, the noise is measured over the bandwidth around that frequency. The bandwidth depends on the windowing function you use (my favorite is hanning) and the time period over which you are sampling. This window helps to reduce the bleeding of a strong signal into the adjacent frequencies. The bleeding occurs since the frequencies measured may not be periodic with the time period of measurements.
To get the power spectral density at a frequency, you divide the voltage measurement squared over the bandwidth.
So you basically take the rms voltage at a frequency from the fft, square it, and then divide over the bandwidth. For noise, this rms voltage should have been averaged over several samples. If you are using dB, then you would take 20*log of the measured rms voltage to get decibels with respect to 1 V^2.
Agilent has a really good app note which may be able to explain this better than me. Let me know if you have any additional questions.
Home
Pages: 1
What does fft actually mean? (Read 4410 times)