I have made a demo board about AD9649, which is just like the design mentioned by CN0272.
When a DC input voltage is given at point 1, a 20MHz cross-talk can be tested at point 2 (ADC’s deferential input), which is shown in figure below.
In my point of view, this “dirty” signal is generated by digital part of AD9649: when 14 bits output pins outputs 0 or 3.3v digital signals with the frequency of 20MHz, a big current will be needed by the power rail of 3.3v; meanwhile, due to the impedance between 3.3v power supply and the digital ground can’t be zero, the instant voltage potential of the ground will be left. As the result, these 20MHz spines are being captured.
I want to know does my comprehension right or not and how to control these spines because I think it is the main factor that limit the resolution of my board.
As shown in figure 2, according to the datasheet of AD9649, it is called sallen-key filter. But I notice that there are some differences between their typologies. Can I use the function: bandwidth = 1/(2πRC) to calculate the bandwidth of the filter? If I series connect an inductors after the register, is there any improvement compared with the RC filter?