Hi,
DIAdem is using several different FFT algorithms, depending on the version of DIAdem and the mathematical function. The basic FFT, witch is calculating the real/imaginary part out of the time signal, is well defined. Each algorithm gives almost the same result. There are only very small differences on the last bits of the float values. The basic FFT algorithm is usually not very large. It is easy to use different functions or libraries depending on the needs. The results have to be the same in all cases.
DIAdem 10.2 is using a new algorithm which is able to do FFT calculations with any number of values, not just the power of two. This algorithm is also used in other NI-products like LabVIEW. A large afford is going into the quality and speed of this algorithm because not only the simple FFT but also a number of additional functions like the sound and vibration toolkit for LabVIEW is depending on this.
The difficult part, when it comes to comparison between DIAdem and other FFT calculations, is always the large number of options. You need to know the number of values, sampling frequency, window function, window correction and the averaging. You can calculate peak, rms or power spectrum and so on. Depending on the options, you can get very different but still correct results.
Crash is much more a question of standards. Starting with sampling frequencies and ending with precise defined algorithms to get a result like the HIC-value witch has to be reproducible and comparable between cars from different vendors all over the world.
As soon as you start to do real world FFT calculations there are so many options and possible ways do get the results you need, that it is impossible to discuss all of them in a documentation of a Software like DIAdem. You could write a series of books about this.
If you have specific questions about the FFT results, you can give it a try and ask on this forum.
Ulrich Bierwisch
