Quote:
|
Originally Posted by Manoel
Quite a coincidence this thread came up; I'm currently studying Fourier Transforms in my Applied Math course, and I have to hand in a paper detailing one of its uses in my future field (Electrical Engineering).
So, I'd like to ask some suggestions to the engineers who use them in their day job - I'm particular interested in Controls applications (I think Chris Hibner can help me on this one, right?  ) |
Here is how I've used Fourier Transforms at my job:
1) To determine frequency content of a set of data that we collected. This helps us to understand how we might want to filter the data. It also helps us to see if there are any unusual characteristics of the data in the frequency domain.
2) To remove unwanted frequency content. We were looking at some data we collected and noticed something strange. In the frequency domain we noticed a huge spike at a certain frequency. We went back to the test setup and noticed that one piece of data acquisition equipment radiated a lot of energy at that frequency and that created the large spike at that frequency. In order to remove the frequency, we took the FFT of the data, removed the spike by setting the frequency domain data to that of the average value of the data surrounding the spike, then took the inverse FFT.
3) Creating complex transfer function models. If we can identify the frequency response of a system emperically (like by using a sine sweep), we can use Fourier convolution to model the behavior of a signal through that system. This can get tricky, though.
One thing to keep in mind when using FFTs: The process of taking a Fourier Transform of a finite-time set of data will introduce false content into the frequency domain (and vice versa when you take the inverse). If you take a straight-up FFT of a time history of data, you are actually superimposing a square wave on top of the data, which becomes a sinc function in the frequency domain. If you understand the effect you can take some care to minimize how it affects your data. Just be sure you don't blindly use Fourier methods without doing a few experiments and looking at the results closely. Once you take the inverse FFT, be sure to look at the beginning of the time-domain data for windowing effects.