What voltage would you expect on an IT system with an undistributed neutral between live and earth?

I am surprised that cable capacitance is that much of a problem is it really that much of an issue in the UK? Also when you talk about an IT earthing arrangement presumably what your really referring to is an isolated supply is this correct?

when you talk about an IT earthing arrangement presumably what your really referring to is an isolated supply is this correct?

Supply isolated (from Earth) or Earthed via a significant impedance (I) but exposed-conductive-parts solidly connected to Earth (T).Have a look at the diagrams in appendix 9 of BS 7671.

A separated supply (like for a bathroom shaver socket) is similar, but a little different - in a way it's better separated from Earth - no deliberate impedance to Earth allowed, the cable lengths are restricted to limit capacitive coupling (100,000Vm - e.g.  400m for a 250V system - less than you'd have in a typical domestic install) and exposed-conductive-parts are not connected to Earth. Separated system usually don't require insulation monitors or other indications of 1st fault, whereas IT systems normally do.

   - Andy.

In a well isolated 50Hz system, i.e not one with mineral insualted wire, or Paxolin or Bakealite insulation, the  capacitance to earth is the dominant mechanism for current flow, apart from the deliberate metering impedance inside any insulation monitor (if present).
The currents are indeed very low, and the impedance is very high by normal electrical load standards.
One metre length of PVC twin and earth 1mm2 as a simple example manages on my meters in excess of 10 gigohms (1E10 ohms ) between the conductors.
But the capacitance is more like 50pf between L and N, and about one and a  half times that between L and E and N and E - there is because there is some capacitance between L and N that is not via E, but it is probably better to think of it as two 75 pF capacitors one L-E one N-E and perhaps another 20-30pF direct L-N.
Now 1 metre is a very short length for a real installation that is not just one shaver socket, so in a real IT installation it is quite possible to reach 1000pF or more to earth - it is after all something between 10 and 20 metres of typical cable. Larger cables have larger cores, but they are also usually spaced further apart, so the core to core capacitance per unit length is only a slowly rising function of cable size.

So how much current? X= 1/(2.pi.F.c) for 50Hz makes 1nF about 3j megohms - and at 230V that has a circulating current of ~ 80microamps - not enough to hurt, but certainly detectable, and factor of a hundred or so more than the resistance of the insulation and for some folk that would be coming into the level of perceptible.

So anything  with more than perhaps a couple of tens of metres of cable has a non-lethal but noticeable 'tingle level current'  just flowing all the time in the self-capacitance of the wiring.
Mike.

And, once the load has an RFI filter with (IIRC) a 0.1uF Y capacitor, you soon start getting 'significant' milliamps.

If you have a 4-port extension lead fitted under your desk, each loaded with electronic equipment (PC, VDU, Printer, ..), it (in some respects) should be 'fixed wiring' because of the potential from a lost earth connection.

At my previous work we had two 4-way extensions in series mounted under the desks and needed an extra earth link between them (every desk!). There had been an incident on an equipment rack (unknown vintage) that had lost it's primary (cable) earth and had (unappreciated) permanent RFI filters on load. Disconnecting the supplementary bonding resulted in sparks!

mind you sparkly equipment can still pass a modern PAT - since 2021 or so the limit for any one appliance is now 5mA see https://www.pat-testing-training.net/articles/iet-code-of-practice-5th-edition.php for a discussion on that.
As you say, a few items that  individually only just pass, plugged in on the same desk or same rack, can soon add up to a single fault lethal level (or in a modern building to at least an RCD tripping level..) in the CPC, despite operating normally.
5mA in the wrong place is actually really !"** painful and as well as a flinch response in a healthy adult, is quite capable of creating little sizzles when a metal contact is made and broken.

Of course, when new equipment is made it is still designed to meet the lower levels of the product standard, but that is now  never compulsorily  tested and filter capacitors damaged by over-voltages and so on will probably not be spotted.
Personally I think the new higher limit, ostensibly to avoid false fails,  now errs on the side of being so loose as not to be worth doing it, and combined with the looser RCD test post 2022 , all conspires to shift the risk balance towards 'more risk', which is perfectly OK ,but we ought to be aware that we have decided to do that, and I suspect folk are not, but that's a purely personal opinion.

Mike

Part of the issue is that the EMC (RFI) requirements for equipment often require some hefty filtering (shunting induced supply noise into the Earth such that the neutral looks to be anti-phase to reduced the L-N aerial effect ;-). Hence the likelihood of excess earth 'leakage' [It's not a leak, it's deliberate] !

Older folks may remember daisy chained power leads through the PC to the VDU. That all stopped when  the RFI regs came into place and they (PC system suppliers) had to some how 'double up' the filtering because they weren't reducing the internal noise. Hence the high L-E capacitance values for the HF noise.

Meanwhile the electrical supply industry was more concerned about LF harmonic noise (for usually obvious reasons when aggregated on 3ph supplies).

Sounds like it's going to run into the same 'overlapping tolerances' problems as the disconnection times discussions ;-)

Broadly agree, and I design stuff for a living that usually has to meet either CE marking or the far tighter defstan EMC levels so I am familiar.

There are tricks  for larger devices, like adding shields between windings that are connected to the primary DC bus not earth, so that the voltage impressed on the transformer inter-winding capacitance to the secondary is only a rectified 50Hz, not 50Hz plus a 100kHz 400Vp-p square wave, as it is mostly the edges of the square wave that sneak through and then need filtering out.

Actually when we know the mains polarity it gets easier, as we can have larger caps N-E and L-N, and not L-E, but rules for both CE and UKCA marking assume we don't know which lead is neutral, mainly because of the legacy of reversible Shucko plugs and confusion even in some countries with polarized sockets. France I'm looking at you.

3 phase is easier, both for this reason, and because with symmetrical filters the 3 phases at least partly cancel.

Mike

Broadly agree, and I design stuff for a living that usually has to meet either CE marking or the far tighter defstan EMC levels so I am familiar.

There are tricks  for larger devices, like adding shields between windings that are connected to the primary DC bus not earth, so that the voltage impressed on the transformer inter-winding capacitance to the secondary is only a rectified 50Hz, not 50Hz plus a 100kHz 400Vp-p square wave, as it is mostly the edges of the square wave that sneak through and then need filtering out.

Actually when we know the mains polarity it gets easier, as we can have larger caps N-E and L-N, and not L-E, but rules for both CE and UKCA marking assume we don't know which lead is neutral, mainly because of the legacy of reversible Shucko plugs and confusion even in some countries with polarized sockets. France I'm looking at you.

3 phase is easier, both for this reason, and because with symmetrical filters the 3 phases at least partly cancel.

Mike