FFT of Downsampled Signal

I have an aperoidic signal that is being sampled at a very high frequency, 3.125 MHz. The signal is sampled for 8 minutes, leaving me with a very large dataset to process (1.3E9 samples). Moreover, I am confident that the data is being oversampled by a large factor so I believe that I can downsample it by a factor of ten. I am chosing to downsample primarily because of computational limits.

Would I best use downsample() or decimate()?

And the real question: How would downsampling/decimating the data by an order of 10 affect the FFT? I would like to have a meaningful frequency axis in the power-spectrum-density plot. Would I just treat my sample frequency as fs/10, 312.5 KHz?

Thanks for any guidance.

Answer 1

[Greg Dionne]

Hi Adam,

If you have significant noise on the signal, then decimate(x,10) or resample(x,1,10) would be preferable. Your resulting sample rate is as you have indicated (312.5 kHz).

I would take a spectrogram on the resulting signal to see how the frequency content changes over time.

Hope this helps!

-Greg