Phase Noise Analyzer - Sampling Frequency vs Spurs

[email protected]

From the classic sampling theory (or digital signal processing), the spurs are theatrically existed. Analysis will be presented here. Matlab scripts have been written to complete some simple simulations. It is not a rigorous script, but match the test results well.

Examples are presented firstly. Test setup: all 4 ports are feeded with a same 10MHz signal (for self-correlation). Let Fs=78MHz and f_DUT=9.99994MHz, the following figure shows the test result along with the simulation result. It is interesting that if the Fs is exactly 78MHz (without deviation) and f_DUT is exactly 10MHz, there are no spurs in the interested frequency range (<1MHz). However, in a real environment, there are always some deviations for the Fs and the DUT as the temperature stability of the DUT & sampling clock is limited.

And following is another example, where Fs is exactly 77.76MHz and f_DUT is exactly 10MHz. It seems that the amplitude of the spurs is well below the noise floor of the device, which is not detectable. However, if f_DUT=9999.94KHz (60Hz deviation), there will be many large spurs again.

Possible solutions: Calculating the spurs and removing them dynamically (including follow the deviation). These algorithms can be implemented based on matlab scripts.

END