Why should analog ground and digital ground be separated in some instrumentation

[nour]

I have noticed that in a lot of instrumentations, the ground of the analog part and the digital part are usually separated and also the connection between them(control and data transfer) are usually separated by optocouplers? Why?

Why does the ground should be separated and why they use optocoupler also between them?

[Ghydda]

I have noticed that in a lot of instrumentations, the ground of the analog part and the digital part are usually separated and also the connection between them(control and data transfer) are usually separated by optocouplers? Why?

Why does the ground should be separated and why they use optocoupler also between them?

The short answer is proberbly: to ensure/control the path of each sections return current.

If two subcircuits share return current path or have crossing return current paths, they will interact.
Usually this is okay for most circuits, but when high precision or low noise must be achieved, then completely separating the return currents is the way to go.

Sendt fra min Nexus 6 med Tapatalk

[tablatronix]

My understanding is I thought it was because digital is high speed and induces noise into the analog stuff, opamps filters etc. so keeping power seperate allows better filtering and isolation. I could be totally wrong.

[Bud]

Ground partitioning may be difficult to figure out when using 'hybrid' devices such as DDS ICs that have both digital and analog circuits inside. I stopped worrying about it after speaking to a engineer at Analog Devices who said they do not use separate grounds in the Lab and when characterizing their products.

[danadak]

Some application notes on this topic -


http://www.cypress.com/documentation/application-notes/an57821-psocr-3-psoc-4-and-psoc-5lp-mixed-signal-circuit-board


http://www.semtech.com/images/datasheet/rf_design_guidelines_semtech.pdf


http://cache.freescale.com/files/analog/doc/app_note/AN3962.pdf


http://www.analog.com/en/analog-dialogue/articles/high-speed-printed-circuit-board-layout.html


https://wiki.analog.com/resources/app-notes/an-1119


Regards, Dana.

[German_EE]

A practical example.

I have on my bench at the moment a phased locked loop that covers 35 MHz to 75 MHz in 0.1 Hz steps. The VCO sensitivity is 3.5 MHz/V or (and this is where it gets interesting)  3.5 Hz per microvolt. So, if I change from 40.0000001 MHz to 40.0 MHz the VCO voltage has just dropped 29 nV, not a lot. If the 3.3V control logic was sharing the same groundplane*** as the VCO the switching noise would modulate the VCO input and I would not have a very good synthesizer.

*** Note that no groundplane is perfect, it has inductance and resistance and therefore has circulating currents.