FFT Peak at 0hz

[bilalskd]

Hey. I am using cDaq and NI 9237 (Bridge module). Where I am measuring strain and performing Frequency analysis on it (FFT). This is known as dynamic stress analysis if I am not mistaken. The issue I am facing is that there is a peak 0 Hz which is relatively large as compared to the peaks at other frequencies. I am using a quarter bridge configuration and the gauge is 350ohm.

How can I get rid of this peak. I have tried a high-pass filter blocking frequencies lower than 3Hz. I have used the butter-worth filter with a order of 35. but this only dampens the the amplitude for that peak and it shifts it to 6 7 HZ.

Another point is that when the filter is applied the 0Hz peak is shifted as mentioned and it's amplitude increases and decreases.

Some feedback will be highly appreciated.

[GerdW]

Hi bilal,

the value at 0Hz is an offset (aka "DC value").

So you have two options:

• remove the offset in your time-based data, then apply the FFT
• first apply the FFT, then set the offset to zero in the frequency-based data

[bilalskd]

How could I remove the DC offset. Could you kindly elaborate more on then and provide with an example if you can.

Thanks A lot for the response.

[altenbach]

---

@bilalskd wrote:

How could I remove the DC offset. Could you kindly elaborate more on then and provide with an example if you can.

---

You can subtract the mean of all points from the data. But why? A 0Hz component is just data and does not influence the other frequency values, you can just ignore it.

[bilalskd]

The reason I wanted to remove it because it was large in amplitude and causing the scale of graph to be affected due to it. 

I could obviously manually scale my graph to see the relavent  frequencies. But just wanted this to he removed and have the pure relavent data.

[GerdW]

Hi bilal,

---

@bilalskd wrote:

But just wanted this to he removed and have the pure relavent data.

---

ReplaceArraySubset with 0 at index and element input will do the trick…

[altenbach]

Often it is also useful to just set the Y scale to logarithmic. Now you can more easily see minor frequency components, even of some frequencies (e.g. 0Hz) are very dominant.