The question and other answers ignore the 2 most important criteria for AC product safety.
1) HIPOT Leakage current to ground ( 100% product testing )
- A much higher voltage is ramped (>3kVdc )between Input (Line AND Neutral shorted ) and ground to verify that leakage current is < 100uA
- actual HIPOT tester level depends on IEC/UL/CSA/etc specs for 60s or 1s test level and DC avoids line filter ac current by design
2) Safety ground continuity ( pre-approved by design )
- the safety ground to chassis must not raise voltage more than 1V with 10A current.
- if there is no ground connection, then insulation system must be redundant by at least 2 methods
There is NO REQUIREMENT for isolation between AC input and DC output with respect to Neutral.
However it may be illogical to superimpose Line on HVDC, it is not forbidden to use neutral as common to the HVDC output. However the following must apply so the design must still pass HIPOT and function as expected.
The hot and neutral wires are interchangeable as far as the equipment is concerned.
ANY EXPOSED METAL MUST BE EARTH GROUNDED.
Inter-System Ground Noise is never solved by a (Galvanic) Isolation transformer.
These have absolutely no affect on this problem.
In some countries and locations, for rural and remote equipment, the HVAC is delivered without a Neutral return wire ( e.g. at 20kV) and relying on the earth for a power conductive path. See Single-wire earth return (SWER)
It does not have to be galvanically isolated from the mains neutral.
But it must be safe to the operator under any condition.
This includes; cable disconnect, power interruption, secondary arc, primary over-voltage, contaminated insulation with any humidity range.
- While low %RH promotes voltage build up on insulation by stray HVDC fields and high %RH promotes leakage current to stray surface paths.
How you make it safe is not determined by galvanic isolation, rather by design of insulation system and protection.
This determined by controlling E fields and HiPot test verification of the system.
- e.g. If 200kV can get in, it can probably get out so plan on ESD/HIPOT testing to all surfaces under worst case dust condtions above your required levels.
A shielded conductive spray and earth bonded is safer inside the plastic case. Then any dielectric breakdown will be safely shunted to earth ground via the inside connections. A Bleeder resistor must have the power dissipation to discharge the cap when the unit is disconncted from primary power when on such that any line filter will not cause a floating high voltage between any exposed surfaces and pin or have some other safety measure.
If you plan on using bushings or insulation, the power industry BIL ratings are for AC and not HVDC, so BIL200 can withstand 200kV lightning but not 200kV AC and much less than 200kV dc. This is why air gaps are rated ~500V to 3kV/mm depending on contamination and surfaces much less.
Plan on having an AM radio tuned off-channel to detect any unexpected and possibly silent static noise discharges from inadequate insulation until to have finished the design verification tests planned. (DVT)
For example I used 1cm thick nylon sheets above an open oil-filled 5MVA transformer tank with two(2) BIL200 rated insulation bushings for 400kV lightning protection in order to test up to 150kVdc. Then static discharges started on the wall 30 meters away with a mounted steel girder for show N tell of insulation components that was electrically floating from earth ground. It then started to tick quick loud like a clock, while I stood a few m away outside the cage. Even though the cage was earth grounded, it too had high voltage on the paint that felt like normal carpet ESD door-knob discharges when brushing against the paint. Even the 100 ton transformer tank paint was electrically charged, when touching it during testsm it gave off static discharge of ~1kV for on every surface of epoxy paint.
To test at 500kV or > 1Gv demands much better gas insulation or drone cleaned surfaces.