I seem to have some trouble understanding the benefits of output randomization, in my case using the LTM9009-14 series. I have to mention that I use it for sampling "slow" waveforms, and am not familiar with SDR applications, SINAD, SFDR, and so on.
Following the description in Digital Output Randomizer for High Speed ADCs | Analog Devices:
At first I just saw the effect that a transition at midscale
0b01111111111111 -> 0b1000000000000
will turn to something much more interference friendly:
However those unfavorable transitions are, in the end, just moved to different locations.
In this case, it's even immediately clear from the reverse operation.
Going just one bit above midscale
0b10000000000000 -> 0b10000000000000
0b10000000000001 -> 0b01111111111111
and we are back with this unfortunate transition.
Can someone explain to me why the randomization is still useful?
Or is this in fact only relevant for SDR applications, because of the way how those transitions are moving from 2^N related positions to other positions?