0
\$\begingroup\$

I'm using an analog switch, ADG1434 to switch some filter components in a circuit. My question is about general best practice for the digital signals going to analog switches. The switch is dual supply so is powered from the analog +/- and uses analog ground. The switch logic input signals however come from the digital part of the circuit, which in my case has its own digital power and digital ground. If I directly connect the digital signals to the switch, then the (very small, maybe < a few nA according to the data sheet) current will flow from my digital side, through to analog ground and return to digital ground elsewhere on the PCB. This seems to me like breaking the rule of trying to keep these signals separate to maximise performance of the analog circuit. Adding optocouplers to each of the logic inputs so they are powered only on the analog side seems to be the only solution I can see.

So what do people generally do? Is there a general best practice for this?

\$\endgroup\$
3
  • \$\begingroup\$ CMOS analog switches have no analog ground. They are purely supply referenced.. The GND pin is basically the digital return pin. \$\endgroup\$
    – tobalt
    Nov 2, 2021 at 6:56
  • \$\begingroup\$ What about the bigger problem of charge injection into the analogue circuitry. You appear to be focussed on the wrong thing. \$\endgroup\$
    – Andy aka
    Nov 2, 2021 at 9:31
  • \$\begingroup\$ Thanks tobalt, that's really interesting to know. Do you have any references for that? I certainly don't understand the internal structure of these devices. If the case, then as you say, GND should just be a digital GND and there should be no current flowing from the digital to the analog sides of the device. \$\endgroup\$
    – JDC
    Nov 3, 2021 at 3:08

1 Answer 1

Reset to default
0
\$\begingroup\$

Running the digital signals through optos still doesn't stop the fact that the the output signal on the opto is still a digital signal and its return currents will still flow between the opto output and around your switch.

What you're supposed to do is divide the digital and analog into separate sides of the IC and route things such that the analog return currents only run under the analog traces and the digital return currents only run under the digital traces. You do not want the digital return currents running through the same areas as the analog return currents or under the analog traces and inducing funny business.

Then if you did that, properly connect the analog and digital reference planes together at some point to hold them at the same potential, but so that no return currents from either side will actually flow through it (which would result in digital return currents flowing in the analog reference plane and under analog traces). This is usually directly under the IC.

But if you actually did it properly then you don't usually need to worry about connecting the analog and digital reference planes at a single point. You just make one big plane and the return currents stay in their own area. One time you might need to actually put a physical slot in the plane with just a bridge between planes under the IC is is if your application is so noise sensitive you cannot noise from "smearing" of the return current path as it flows through the plane from bleeding into the other side (think about how water flowing a floor doesn't just flow in a straight perfect line from A to B. It will more or less follow path centered around the straight line from A to B but spill, spread, and smear out over that line as it flows).

http://www.hottconsultants.com/pdf_files/june2001pcd_mixedsignal.pdf

\$\endgroup\$
8
  • \$\begingroup\$ I suppose the idea is that with an opto-isolator the signal on the analog side won't have very much noise coupled from the digital side. \$\endgroup\$
    – user253751
    Nov 2, 2021 at 10:12
  • \$\begingroup\$ @user253751 It does prevent switching noise from other things the MCU is doing from coupling into the lines connected to the switch, but does not do anything about the opto itself switching. This is similar to why you might run the SPI bus going to an ADC through a transparent latch first so that when you are communicating with something else on the bus, the noise from the SPI lineS toggling doesn't reach the ADC pins even if it is ALREADY ignoring them due to chip select. \$\endgroup\$
    – DKNguyen
    Nov 2, 2021 at 13:17
  • \$\begingroup\$ well, I assume that a transient when switching an analog switch is hardly unexpected \$\endgroup\$
    – user253751
    Nov 2, 2021 at 13:31
  • \$\begingroup\$ @user253751 Let's try not to stray talking about both your core problem (separate analog/digital ground, return current path) and the proposed but inappropriate solution actually does and where you would use it (optos). Optos do not in and of themselves address the separate ground analog/digital and return current path problem you ask about. You can still route the opto output and return to overlap your analog traces and return to still get the same issues. \$\endgroup\$
    – DKNguyen
    Nov 2, 2021 at 13:40
  • \$\begingroup\$ my core problem...? \$\endgroup\$
    – user253751
    Nov 2, 2021 at 13:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.