AC->DC supply, how to deal with chassis ground being connected to digital ground?

Question

I have an external triple-output AC-DC supply with the following output pinout:

On the power supply, "ground" (pin 2) is internally connected to the ground wire of the mains plug. I'll refer to it as "earth ground" (it's "AC GND" in my diagrams below).

Current Configuration

With the PCBs that I already had built, "common" (pin 1) is connected directly to the board-sized ground plane; let's call it "digital ground". There are various connectors on the board, such as a USB jack and an analog input connector. The connector shields/shrouds are connected to the digital ground plane where they are soldered to the PCB, and the shields/shrouds are also connected to the chassis/enclosure. The image below shows the configuration:

The device is used for measuring very small currents. The DC power supply's +12V and -12V outputs drive sensitive amplifier, DAC, and ADC circuits.

My concern is that since earth ground is left floating, the enclosure could theoretically become energized due to a fault. The voltages in my circuit max out at +12V / -12V, but does that even matter from a safety standpoint?

Possible fix for existing boards

I've read this answer to the question Should chassis ground be attached to digital ground?, and the way I understand it, I'll be OK if I just connect earth ground to the chassis at the power input connector, since my chassis is connected to the shrouds of all the input connectors already. Without making new boards, that's really the only thing I think can do right now anyway. Could this turn my device into an unintentional radiator since digital ground is connected to the shrouds too? Could it also put more noise in my power? Any other serious concerns?

What to do with the next version

Ideally, I think the grounds should be hooked up like this, so that AC ground is isolated from the rest of the board, and there is no way for the chassis to pass current except through external faults or a serious internal fault, but the current would go to earth. I could change the layout so that the next batch of boards are hooked up like the diagram below. Do I have the right idea?

Plastic Enclosure

I have one more issue that complicates things further... This device is intended to be produced in two different configurations: in the primary configuration, the enclosure is aluminum. However, in the other configuration, there is a plastic enclosure. This means the only exposed conductive parts are the connector shrouds. Of course I'd like to use the same board if possible.

What's the safest way to deal with the exposed conductive connector shrouds? Should I leave them floating, or just connect them to digital ground? The plastic enclosure does have an internal conductive coating, but I don't think I can make a reliable connection to it (let's assume I can't).

_{I have already read this other question that is very similar, but I have a different issue since I'm working with an existing board that already connects chassis ground and digital ground.}

Answer 1

Any long cables with RF clocks/data will radiate noise and when connected to AC ground could also radiate via AC cable from path ground loops.

Smart idea is make generous use of common mode chokes for interface cables and DC outputs to raise impedance of the common mode signals from AC ground.

Beware that Hipot testing of AC/DC PSU may cause failure on secondary port when grounded. This is due to dielectric breakdown of Hipot from primary to secondary. Floating secondary is easier to protect if there is capacitive leakage and tester uses fast rising input. In this case modify Hipot tester to avoid discharge surge that would destruct PSU by using a current limiting series resistor.

Leakage can be minimized with prudent gaps, quality caps, and moisture protection on conductors. (moist dust causes leakage)