FFT resolution and frequency spectrum

[John Sigurdson]

David,

Starting with a pressure waveform in psi with 30 ms of data recorded at 1 MHz (30,000 data points). I lable the X axis as microseconds and Y axis as psi.

When I do an FFT of this signal, it produces a FFT plot of magnitudes from 0 to 500 KHz with one point (magnitude) about every 30 Hz ( n(1) = 16,385 ).

Q: Is there any way to increase the number of points so as to get 1 Hz resolution for waveforms sampled at 1 MHz? That is, n(1) = 500K. That would allow calculation of a standard 1 Hz frequency spectrum and, with added calculations, subsequent energy spectral densities etc. Or, is there a way to finesse that with existing settings?

John Sigurdson

[DPlotAdmin]

Q: Is there any way to increase the number of points so as to get 1 Hz resolution for waveforms sampled at 1 MHz? That is, n(1) = 500K. That would allow calculation of a standard 1 Hz frequency spectrum and, with added calculations, subsequent energy spectral densities etc. Or, is there a way to finesse that with existing settings?

Only by increasing the number of points in your input. This isn't an arbitrary decision made by DPlot; it's how an FFT works. The frequency interval in the FFT will always be 1/(2*dt*(Nfft-1)), where Nfft = number of points in the FFT, always half the number of points in your input plus 1. For dt=1 microsecond and 30,000 input values, the frequency interval is 1/0.03 Hz = 33.33 Hz.

And after typing all of that I realized it was an overly complicated way of saying that to get a 1Hz interval, you need 1 second's worth of input. 1/(2 * 1E-6 * 5E+5) = 1.

[John Sigurdson]

Thanks, again, David

I should have known...