I am still struggeling to understand the concepts of SDR. I studied the source code auf sdr# and “Gnuradio rtl_fm” to get an idea on the practical realization. Currently, I am using a DVBT-USB stick and started with a c#-project to visualize the spectrum of the FM radio frequency range. The program works quite well. With a sampling rate 0f 2048 KHz I can visualize 2 MHz chunks. I use the FFTW3 to calculated Fourier analysis with 2048 samples and get nice peaks of the different radio stations. However, when I analyze the frequencies of the peaks I found that they are off the known frequency by several KHz. I assumed the problem is caused somehow by the fft algorithm und used the Lomont fft (http://www.lomont.org/Software/Misc/FFT/LomontFFT.html). However, this did not solve the problem. The peaks obtained by the Lomont fft differed also from the known frequencies, although the differences were disparate from those using the FFTW3. When I feed the signal into the Gnuradio companion fft-sink the peaks correspond to the correct frequencies.
Is there a correction factor necessary?
