Y capacitor provisioning when the output is referenced to mains neutral?

Question

I am designing a device that uses a "standard" board-mounted AC/DC switching supply module (on the order of 20-30W), along with a battery backup, to power itself and an attached device, both of which have their circuit grounds referred to mains neutral. (It's related to the situation here, and indeed the attached device in this question is a hardwired smoke alarm. Isolating the signal connection that would otherwise be referenced to mains would have serious downstream knock-on effects in terms of usability and functional compliance, so I'd rather not do that.)

For testing purposes, I have set it up so I can operate the device effectively single-station from an isolated "surrogate" DC supply. However, I also want to leave provisions in my initial layout for additional EMC components, as the device in question is subject to a product standard (UL 217 in the USA) that imposes additional EMC conformance requirements (everything from additional transient/surge testing to "ad hoc" ISM-band radiated immunity) above and beyond normal FCC/CISPR Class B emission standards. (It also seems that some supply modules aren't as great about meeting the full range of EMC standards without a bit of external help as others are.)

So far, I have left room for a SMT fuse, a big integrated GDT/MOV combo surge suppressor (which could be supplemented with an in-board spark gap if need be), an inrush limiter (NTC thermistor), a whomping large X capacitor, and a chunky common mode choke. Normally, though, you'd have Y capacitors in this as well, either referenced to mains earth, or to the supply output, as depicted below. However, I'm not sure how one'd configure the Y capacitors to render them effective, given that there's a connection from mains neutral to the negative side of the mains neutral.

Out of C2, C3, and C4, which parts should I be leaving space for, and which are useless or counterproductive in my scenario? Also, is my "T" topology generally connect, or should I be connecting the DC side at a different point and/or bridging out one of the caps in question?

^{simulate this circuit – Schematic created using CircuitLab}

Answer 1

Correct, with a three-wire connection, you have a common node, so a common-mode filter is moot. You do need to wire it such that this remains true. For example, it is better to make three connections at one location on a board, rather than pairs of connections on opposite ends of the board.

^{simulate this circuit – Schematic created using CircuitLab}

That is, contrast these cases, where the dotted line represents a physical PCB, which can have some ground loop voltage VCM across it.

If we simply place filter components such that the three terminals are pointlike together (that is, minimal distance between them, relative to the frequencies and impedances of interest), we minimize noise at the input, output, and between them.

^{simulate this circuit}

I'd draw the capacitors scrunched in if I could, but I'm stuck with the library symbols, so imagine the folded-over wires to them implying they're actually right in there with the parallel wires and inductors.

Note that inductors have high impedance at high frequency, so really the only contiguous path along here is the middle / neutral wire, and the capacitors serve to join the three wires together at HF (approximating a supernode) above and below the inductors. Thus greatly attenuating high frequencies.

Note that there could still be a sneak path, from self-capacitance of the PSU through air to the surroundings. This limits ultimate performance of a naked assembly. The solution is to add a shield, tied with the ground point at the filter -- preferably, the wires penetrate the shield at the same ground point. Of course a metal chassis wouldn't be tied to neutral here, but a Y-type capacitor would do the job. It's RF grounding that I'm speaking of here, of course.