I am using a C# code to capture data from the AD9269 evaluation board through the HSC-ADC-EVALCZ board, and I've noticed that the FFT of the captured samples results in a very different spectrum tan when using Visual Analog.
In the attached MATLAB screenshot, I've input two sines (one at 3 MHz and one at 5 MHz). The ADC is sampling at 20 MS/s (I've tried with both the internal clock of the evaluation board, and using an external one). As you can see, although the most powerful component is correctly at 3 MHz, plenty of spurs appear, specially at multiples of the input frequency. In the 5 MHz it's not so obvious, because the next multiple would be 10 MHz, which is the Nyquist limit (but you can still make out a component in the 10 MHz mark). This happens for all other frequencies inside the Nyquist band.
This does not occur when using Visual Analog, where I use the sample AD9269 "FFT" canvas, as you can see from the attached screenshots. The spectrum is a bit noisy, and some spurs occur, due to the generator used for the input. But it is nowhere near as bad as the one captured with C#.
I've also captured the code with MATLAB, and the result is the same as with C#. In all 3 cases (C#, Visual Analog, MATLAB) I am using the same .bin file to configure the board.
I would guess that something is not being correcty configured in the third-party codes. Does anybody have any idea why this could be happening?