Offline SMPS's isolated secondary must be earthed?

[Faringdon]

Supposing you have an isolated offline 20W SMPS on a PCB in an equipment. The equipment has metal walls in its enclosure. -As such, im sure you would agree that this metal case must be earthed.
Also, on the offline SMPS PCB there is also some (secondary side) analog circuitry with microcontroller etc which is supplied by the offline SMPS.

Would you agree that the secondary GND of the SMPS must be connected to earth to reduce noise problems generally?

[Siwastaja]

No, it doesn't need to be. The proof is in gazillion of supplies on the market, all of proper quality and certified, which do not do this.

Some do, some don't.

[penfold]

With regard to noise, not really, but... Most SMPS transformers will happily couple through a fair amount of switching noise via the interwinding capacitance -- if the secondary is high impedance wrt "some defined potential" then you'll just have a secondary whose 0V is bouncing up and down, not necessarily a problem in itself. If the secondary load is exposed to radiate or you have more "system" connected which allows it a return path to mains (live, neutral or mains-earth) then you'll get a noise current flowing or radiated emissions -- both cause problems. The input to the SMPS will presumably be connected to "Earth" via a Y-cap, so yeah, maybe connecting secondary to earth could help in proving a return path for noise, though it can be done via a Y-cap to prevent some of the woes of earthed secondaries. I would expect the PSU to already have appropriate Y-caps at the optimum position, so if you do have a noise issue at its output, the effectiveness of any connection to earth (direct or Y-cap) at any significant distance from the transformer would be significantly less than one that's "part of the supply", so CM-choke + additonal Y-cap/direct earth connection would be better.

For safety, consult your requisite standard. Normally only need to keep the voltage within some defined limit wrt chassis earth, connecting it to mains earth is one means of doing that but not the only one.

[Gyro]

Supposing you have an isolated offline 20W SMPS on a PCB in an equipment. The equipment has metal walls in its enclosure. -As such, im sure you would agree that this metal case must be earthed.
Also, on the offline SMPS PCB there is also some (secondary side) analog circuitry with microcontroller etc which is supplied by the offline SMPS.

Would you agree that the secondary GND of the SMPS must be connected to earth to reduce noise problems generally?

Again with the over generalisations treez.

No I wouldn't agree. I have designed several Class II metal cased consumer A/V products containing SMPSs which passed EMC (and LVD) certification with no problems, and as Siwastaja points out, there are huge numbers of such products in use and in the market.

[Faringdon]

Thanks,
Our product is supplied by a cable with Live, neutral and earth, and so it is Class I.
If we do not connect the isolated secondary GND to earth, then it will be floating, which is going to be bad for emissions, would you agree?
[Also, i believe we would all agree that (though it wouldnt happen for safety reasons), if the primary and secondary of a offline flyback converter were connected together close to the transformer, then that provides a vast reduction in EMI problems.]
Its a well know fact that non isolated converters provide far less EMC problems in general, than isolated converters.

When its deemed that the secondary GND must be earthed, one then could look at how to go about doing that….
Would you agree, It needs to be done in such a way that there are no earth ground loops?, or at least, a minimal number of earth ground loops….as follows…..

..So Earth enters the product through the mains cable…the mains cable joins the mains input connector on the PCB at the offline SMPS’s AC input.….so we should be sure there are no other earth wires coming to the product other than that. This then means that earth has to be gotten from the primary side of the offline SMPS, over to the secondary side….
..the simplest way to do this is to simply have a thick, well insulated  earth wire jumping over from primary side to secondary side.
The metal enclosure also needs to be earthed (for safety and EMI reasons). We don’t want the connection of the chassis to earth ground to form a big earth ground loop…..therefore, we have to make the chassis connection from the bit of earth at the offline SMPS input  earth connection.
Does this sound like a plan?

[penfold]

...
If we do not connect the isolated secondary GND to earth, then it will be floating, which is going to be bad for emissions, would you agree?
[Also, i believe we would all agree that (though it wouldnt happen for safety reasons), if the primary and secondary of a offline flyback converter were connected together close to the transformer, then that provides a vast reduction in EMI problems.]
...

What do you think the purpose of a y-cap is?

The problematic noise is not at LF, so why would a direct connection to earth be any better than a y-cap between primary and secondary?

The noise is being injected from the rectified mains side to the secondary, the optimum solution to noise is to provide a HF short as close as possible from the secondary to the source of the noise (y-cap). Earth is not going to provide the lowest impedance return path since the inductance of the loop from an earth located near the secondary to the rectified mains side is going to be physically quite large.

Why don't you just try it? Sure it can be an expensive approach, but so is faffing around speculating.

[Faringdon]

..thanks yes i agree on Y caps...though as you know, Y caps have parasitic inductance...which means the usual thru hole ones arent too great at the high hundreds of MHz. (i appreciate that conducted emissions tests  only go up to 30MHz, but as you know, common mode noise is the causator of high frequency radiated emissions which is tested up to 1GHz.)

Anyway, we both agree that a y cap  is needed across the  isolation barrier.

The crux point is that of earthing the secondary ground....and if it should be done, then how do you best do it?...as above (post #4)?

The alternative to not earthing secondary ground, is to leave it floating, which seems like a bad idea from an emissions viewpoint. Th eproduct is in a metal case, but its going to be difficult to completely avoid having openings in the case, which even if small, can result in radiated EMC problems. Even if we dont have an opening now, a future potential modification may mean having one later.

[penfold]

...though as you know, Y caps have parasitic inductance...

Agreed, but given their position, that relatively small inductance will be significantly lower than the path via earth.

Since the earth point isn't going to provide you with any magic low impedance path to short out the noise, you can make it relatively lower by placing a CM choke in the secondary before the point at which your secondary 0V is connected to earth, which will give you potential divider effect and most of the voltage will be dropped over the CM choke and not over the earth impedance. Whilst you could follow that up with "as you know, inductors have parasitic capacitance etc."... that is why you select the filter components appropriately around the problem at hand and ensure that it provides the best attenuation at your problem frequencies.

...
The alternative to not earthing secondary ground, is to leave it floating, which seems like a bad idea from an emissions viewpoint....

TBH, connecting the output to earth "just 'cos" seems like a worse idea to me.

[madires]

...
The alternative to not earthing secondary ground, is to leave it floating, which seems like a bad idea from an emissions viewpoint....

TBH, connecting the output to earth "just 'cos" seems like a worse idea to me.

I agree, it depends on the specifics of the design/device and its environment. Besides EMI there can be other reasons like safety or preventing a tingle from the EMI supression cap between primary and secondary side. It can be also the other way around, i.e. you want the output to float. In some cases earthing the enclosure and/or output even introduces noise into the circuit.