We use TIA (current to voltage) very low noise and we want to get 20 ENOB. As I see in table 26 of the AD7770 datasheet we need to use 2000 SPS to get 19.77. But if we go to 8000SPS we get 18.76 which is only 1 bit less, so if I use oversample of 2^1=2 we will get again ENOB=19.76 with 4000SPS. So it mean we get the some dynamic range in better speed.
So my question is what prefer. To use over sampling when I do the decimation or go on lower freq where the device give me the result?