Please post your block diagram so we can respond. However, I do have a couple of questions that may help you out.
- Are you absolutely sure your data peak amplitude is not
changing? If you are acquiring data using a digitizer, the
amplitude will almost certainly change at each iteration due to
noise. How much is your value changing? 10%? 1%? 0.1%? Is
this in the noise spec of the digitizer?
- In the ideal world, an FFT would not change when you changed the
length of DC components at the end. However, FFTs are calculated
using finite precision, so roundoff errors do occur. How much
they effect you depends on the algorithm. Once again, how much
variation are you seeing?
Post your code, sample data (if any), and your signal variation and we will try to help more.