Why the frequency offset is unstable ?

[HustLiliAn]

Hi, everyone  

We are very confusing about the frequency offset between two different USRP 2920, it is unstable. The transmission wave is a single cosin wave cos(2*pi*fc*t)  in the I path, fc is the carrier frequency of the transmitter.  The received signal of I path should be cos(2*pi*(fc-fcr)*t), fcr is the carrier frequency of the receiver. We set fc and fcr the same at both transmitter and receiver 400MHz,   and fc-fcr is the frequency offset we want to estimate by the FFT algorithm .  The IQ rate is 500k, and each time we capture 500000 points to estimate the frequency offset by the FFT algorithm . We wish the frequency offset is stable, however , it is very unstable .   The following picture illustrate the concrete situation . And the std value is 34.6694, the mean is 112.4722 .   Note we do not run the program again and again to do the experiments. The total estimates is 4066 in a single program running.  Is the unstable phenomenon normal for USRP ? 

 

[ErikL]

Great question. Please check out this white paper on that exact topic.

http://www.ni.com/tutorial/14705/en/