If you can also read articles which exactly recommend the opposite and mention that all other solutions are false where is the sense of such an article then?
The author even contradicts to what is written in other datasheets of converters from analog devices, which was written by another heroes of anaolg devices.


I don't expect cookbook solutions, but that the author is not telling only half the truth. A bit more deepness and some more helpful explanations I should can expect from a heroe of analog devices.


Yes, but if you actually know the basics, then there's no need for such an article. Articles are to learn, and therefore they need a minimum of details and explanations.

Again, this solution will only work with sophisticated line filtering of digital signals, unless you use only the lowest precision converters. Without this important measure the solution will fail and is expected to fail especially from people who understand the EM theory very well. Not because they are anxious or unskilled but just because they have experienced right this: If you access the converter via a fast parallel bus, then you will couple so much digital noise onto the analog ground, that you will never be able to save the precision of a 16bit converter.

The author contradicts himself: First he says, that ground noise is so high, that analog and digital ground pins must be connected right at the package. But then he assumes that the analog ground is free of ground noise and is not contaminated by the digital ground return currents flowing along it. That's plain bogus! If you do what the author recommends, then you convert the analog ground plane into a digital one, just because of the digital ground return currents which now are flowing across analog ground when the converter is being accessed by the micro. More, you also inject all the nasty common mode noise into the analog ground plane.

Kai