I am trying to measure the water velocity of a stream along a channel using an electromagnetic flow meter at a specific location and depth. When I converted the data into frequency domain as power spectral density (PSD), there was a spike at the frequency of 25 Hz.
Out of curiosity, I tried to make a recording using a bucket of still water, and similarly there was a spike at 25 Hz too, although at a lower magnitude (please see the plot below).
Also, the PSD starts quite highly from the lowest frequency, which slowly decreases. I was kind of expecting a simple flat plot with some noise, but maybe I was wrong.
In the plot below, the red line is the recorded data from the water stream (Ch2 (Original)), while the green one is the recorded data from the still water (Noise).
My questions are:
- Are the spike and the general decreasing trend expected? If not, what might cause them? I suspected the old uncalibrated device might introduce noise in the form of a spike, but I am not sure.
- I would like to get the "peaks" information to find the water vortex frequencies, but this will definitely include that spike frequency. How do I remove that spike such that I can get the actual peaks? I tried to use notch filter but depending on the parameters, it either didn't eliminate the spike enough, or too much...it was quite difficult to configure.
Any suggestion and feedback is welcome. Thank you so much!
UPDATE
The sampling frequency is 100 Hz.
This is the specification document to the flow meter: Kenek VM2201・VMT2-200-04P
This is some information about the data logger: Kenek AD1016
The time-domain plots are added below. They are all from the same data, only one plot is zoomed in from the previous one. I think the labels on the x axis is wrong. They are just numbers in here.
