Found it - on Martin Clark's excellent 'acoustica' site.
Can't link direct as the site uses frames but here is the text. Sorry it is rather long but serves as a wake up call to those inclined to dabble.
-----------------------------------------------------------------------
One apparent alternative in order to provide a local earth would be to connect the new earth to the outlets supplying the system, and disconnecting the service earth. This is idea is worth exploring further only to highlight the risks it brings. The fundamental problem with DIY earth rods is simple: it's difficult to get even close to the service earth impedance (usually c. 0.2 - 0.5 Ohms)- and the stake's impedance will vary with soil type, season, rainfall etc. One poster on the Naim Forum connected a digital multimeter between his house earth and a new earth stake in the garden:
AC was more of a surprise (the numbers weren't this high
when I first installed). PD is about 1.5V, and current about 150mA.
I presume this must be 50Hz, but there's no way for me to measure it.
This earth spike equates to ~10ohm impedance: rather good, actually, for a single spike. If it were 'perfect', close to 0 ohms, the measured current would be correspondingly huge; V=IR and all that. Now imagine this earth stake is collecting fault current from some sudden fault in your audio system; with your earth stake at, say, 25ohms - more typical for a 4 foot rod - the current that will flow will be ~10A at UK mains voltage. This may NOT be enough to blow the faulty equipment's fuse, the basic intention of safety earthing. Consider the implications well. This is why TT systems must use of RCDs set to trip at 30mA leakage and not fuses in the consumer unit.
Finally, if you do disconnect the provided earth on a PME supply and substitute your own there is a real risk of electrocution during a fault in the rest of your house or, theoretically, those of your neighbours too. No wire or patch of earth has zero impedance, so fault current flowing to earth elsewhere will raise all bonded metalwork (think radiators and pipework) to a significant potential above your own earth spike. If, for some bizarre reason, you happened to be touching, say, the radiator while you fiddle with your hifi during such a fault there's a finite risk of electrocution. If you mistakenly happen to ground part of your kit to the radiator, whilst still relying on your own independent earth under such conditions, well you better budget for replacing at least the speakers... For information, 2 meters is considered sufficient separation between bonded items for certain classes of separately-derived earth bondings under IEE regulations - presumably because it would require a determined effort to bridge the gap!
In passing, and here because it belongs nowhere else really, a caveat that applies to all PME system users: do not connect earth and neutral together directly at any point, eg using a 'modified' plug, thinking that if they are bonded once, more must be better. The reason is fairly subtle: the earth wire exists as a separate wire inside your house for fault currents only - an escape path. If it was connected to the neutral it woud effectively become exposed to normal load current. If your neutral develops a fault, everything would continue to operate as normal except that all the items normally bonded to earth would float up to some indeterminate potential, possibly dangerously high, because the earth is now a service conductor. Think about that for a bit...
Note again that all this scary stuff refers only to PME systems. Where PME is not the norm locally, the rules are different again. If you have an electrical supply which is earthed to a distribution company spike near the point it enters your house ('TT' - the norm in many non-suburban locations) there is no problem, and multiple earth spikes - bonded as required by your local electrical codes of practice - will indeed improve matters by lowering the earth's apparent impedance. Multiple earth spikes should be spaced at least 2x length apart for lowest impedance.
So if it is all so problematic, why bother?
OK scary stuff over, this one is simple: we've heard the difference a new stake can make and it can be remarkably unsubtle. I (martin) have a strong suspicion I can't as yet test that the reason(s) is/are less to do with 'low resistance' and more to do with 'low impedance'. I'm not talking RF here - just a few feet of connecting wire has too much inductance for that - but I suspect that at audio frequencies, particularly in the midrange, a local earth stake can provide a lower inductance route for noise currents than can the mains wiring; and that this is the crucial difference. Yes there are noise voltages at these frequencies created by audio gear - the greatest source is rectification (diodes) in power supplies, particularly in high power amplifiers, which generate all kinds of harmonics. Compare that with the conventional view of what the earth connection does:
I still don't see why these figures should affect sound quality -
the current flow comes into the house through the supply,
travels between two adjacent connections on the earth
connection bar, and straight out to the earth rod.
Well yes, if the current were constant, and at one dominant frequency; but broadly it is neither. The earth conductor, although intended only for safety use, also sees the leakage current from many, many items of consumer electronics with EMC-compliance filters; your house's wiring, including the earth, may even act as an aerial if you live close to powerful radio users - CB sets, radio stations etc. Since there is obviously some impedance involved, anytime the current or relative magnitudes of harmonics change, the apparent voltage and spectral content of this noise current will as well - and that could provide some odd effects for a system would 'see' a changing reference. Remember these currents will appear as 'common-mode' noise to the system and so stray capacitances, such as exists between the primary and secondary of all transformers, can be a source of coupling into the audio circuitry; there are certainly some subtle effects at work. Some of these ponderings come back to the mains noise investigation/musings and the measurements I've been playing with since. Whilst the jury is still out, I've been learning some rather interesting things (to me anyway!) as to how equipment interacts with the mains. Drawing current at 50Hz is just a part of it in terms of sonic impact; Andy Weekes commented:
I strongly suspect that it's the RF side of things where the
improvement comes - it's a complex thing to model, and I
suspect in particular installations that certain earth cable
lengths may sound better than others, dependant upon
local RF levels and frequencies.
It's such a big improvement though, and surprisingly it's rhythmically
better rather than just quieter in a HiFi sense - weird.
I suspect that the ease with which low-noise, non-degenerated
bipolar transistor amp stages can be pushed into non-linearity by
tiny amounts (relatively) of HF noise, has something to do with it
too, if my recent PSU work is anything to go by. Just small
amounts of unwanted noise can cause severe ill-effects.
We'll leave it there for now...just two things:
Play with your electricity supply and you are on your own. Do we need to say this ?
From 01 Jan 05 in the UK Electrical works such as this need to be signed-off by a suitably-qualified person or approved by your local Building control inspectorate, following the introduction of the new Approved Document Part P component of the Building Regulations.
Further reading:
- UK Building Regulations, Approved Documents Part P
- Leaflet EMC 07: Protective multiple earthing on the RSGB Website.
- Regulations for Electrical Installations (Edition 16) & lEE On-site Guide, available at good bookshops, your local library or via interlibrary loan.
- BS7430:1998 (Code of Practice for Earthing) but at £116 for 86 pages it's probably worth asking the library first..
only a pm away... or any other interested pfmers locally can have a listen too... 
