Sampling can also be thought of as a multiplication of an continuous signal with a Dirac comb so that also plays a role. By cutting out a piece of signal for your analysis your essentially multiplying by a rectangular window function which will convolute with the signal one. With Fourier transforms a multiplication of two terms in the time domain is equivalent to the convolution of the spectra of both terms in the frequency domain. Since you are not actually adding any useful data it’s more of an interpolation though.Įdit: A more complete answer is that you are also suffering from spectral leakage. If that is not feasible in your application you could try to increase the apparent resolution by zero padding the signal. ![]() ![]() The more data you have the greater your resolution of the spectrum becomes. The easiest fix is to just collect data over a longer timespan. Based on your location, we recommend that you select. As you said you know it should be 1.0 but there is no „point“ there. Choose a web site to get translated content where available and see local events and offers. You basically have discretization errors. If I’m understanding you correctly you are testing with a signal of a known frequency and the peak frequency you get from the fft is slightly off so when you try to reconstruct the original signal by inverting the fft you get too large of a difference to the original signal? ![]() Well this doesn’t really have anything to do with linspace itself does it? It’s more of an issue with DFT.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |