Analog and Digital ground... straight from horse's mouth.

I've seen question on this board about AGND and DGND and the most often answer that pops-up is "star grounding."
Personally, I don't use "star grounding"... because it's impossible to implement on a complex PCB.
On my low current design, I just use a "ground plane" and combine both my AGND and DGND.

But here's the latest paper from TI, and the answers may surprise you.

the answers may surprise you.

Go on then what was the surprise?

From the title I though it might be something to do with burgers.

Vasquo, thanks for running down the link. Good info is always useful, :-). Will be
interested, along with G_M, on hearing what the surprises are.

Personally, I don't use "star grounding"... because it's impossible to implement on a complex PCB.

Also, I don't necessarily buy this statement. The first step is to partition the areas of
the pcb used for power supply, digital circuitry, motor circuitry [if used], and analog
circuitry. Then, I think, you can still use star-grounding on most of it, although you
might play with where the analog ground is connected over.

Also, both PIC and ATmega chips use a compromise, in that they specify using a
low-pass filter on the AVCC pin, even if there is no specific AGND pin on the chip.

Also, I imagine the exact best point for making the analog ground connection is
mainly important for cases where ADC resolution is 12-bits and more, although I have
no specific evidence for this.

As a follower of the star-ground principle, I agree with oric_dan. It's much more reasonable to segment larger projects into "sections" with their own ground planes or what-have-you, and route those back to a defined star point. If you're using a two-layer board, you can route power and ground leads next to each other, or on top and bottom via the same path to minimize inductance and EMI.

I was mildly surprised by their recommendation to use the analog ground for low-current digital sections on mixed ICs, but it makes sense in light of the layout guidelines they present. Looking forward to article 2.

This makes me rethink my last project, where I had a control board with analog audio circuits and an AVR controller. The board I/O header had PSU ground and audio ground. I probably should've used the same grounds for audio and the analog IC decoupling caps, and used the second ground strictly for the digital parts. That might fix a noise problem I'm having...

This statement was a surprise to me

Note that the pin names, AGND and
DGND, refer to what’s going on inside the component and
do not necessarily imply what one should do with the
grounds externally.

I've made a high gain small signal amp (2000x voltage gain), using a mixed analog/digital chip and yeah I just combined the AGND and DGND in my board.... it was a prototype board and curious to see what would happen. Happy to say, it's very low noise and didn't see any adverse effects. Again, maybe because it's only low current digital. So I'm glad to read this "the AGND and DGND pins should be joined together externally to the same low impedance ground plane with minimum lead lengths"... sort of a validation. How I did it was just fine.

For high current digital, partitioning and "moating" will probably be essential. My question now is "when does it become high current?" Is it 300mA, 500mA, 1Amp? The article didn't specify.

On that same project, I have a separate control/logic board with the AVR, GLCD, encoders, switches, a local 3V3 regulator, etc.... and the logic board and (2) daughter boards, are all hooked up via ribbon cables. The ribbon carries power, various control signals and the SPI data.

For high current digital, partitioning and "moating" will probably be essential. My question now is "when does it become high current?" Is it 300mA, 500mA, 1Amp? The article didn't specify.

For me, it doesn't matter. Always partition analog from digital, and motor/TO-220
MOSFETs from everything else. You just have to envision the board that way from
the getgo.

Even so for Arduino boards. To me, having the Aref trace going all the way across the
bd to the digital side, crossing lots of digital traces, and on a header with the digital pins
was a foobar.