Re signal ground vs. chassis ground earth:

Under the provision that my house's mains installation is ok, the "mystery" of floating ground is solved. I have standard (3 wires) mains installation with earthed protective ground; I tried several outlets on different floors and always got the same result in what I tried out today. If someone else could reproduce it, I'd be grateful.

Just connect a transformer with 2 * 24V secondary to mains (doesn't matter if with or without mains filter), with a centertap formed on secondaries with no load at all.
Measure above 50V AC with a DMM between mains earth (i.e.: chassis ground) and centertap as well as between earth and any of the two secondary rails.

it doesn't even matter if you have formed a centertap on the secondaries or not; you could leave all secondary wires open.

this should also explain why each ungrounded source (with floating signal ground) connected to the amp adds about another 10V AC to this potential.

Since centertap is connected to signal ground on the amp PCB, but was not connected to chassis, it is no wonder that one gets a hazzard if you touch the amp's chassis (earth) with one hand and RCA connector which is insulated from chassis with the other hand. (same or even stronger hazzard if you touch chasis and RCA connector with only one finger at the same time.)

Conclusion: either 1) don't earth chassis (and connect signal ground to chassis or not), or 2) earth chassis as well as signal-ground.

The first solution needs so-called double insulation of the mains part; the second could produce ground loops.

If however, in the first solution, any of the input or output devices has signal-ground earthed, all effort to have a floating signal ground in the amp is obsolete; this might even produce a strange flow of potential from the amp to the connected devices via signal-ground - with the consequence of hazzard.

Edited by user Saturday, June 30, 2007 6:15:24 AM(UTC)  | Reason: Not specified

the problem with the potential can be solved with a 100 ohm resistor as shown in this article La Ronfle (in french, "La ronfle" means "The hum"), referenced from this page (in french).

the illustrations in the article should be self-explaining even without understanding french.
(for printing the article I had to convert the DOC into PDF first, using the free PDFcreator tool.)

some of the problemes discussed in the article don't materialize with the RevC design since the bridge and capacitors are on the monoblocks. in addition, if you use a passive pre-amp like me, there is no pre-amp signal mass, since both, Darwin selector and Joshua attenuator use the seperate PSU only for switching the relays and have no contact with any signal mass.

in case your amp hums if you connect chassis-earth and signal-ground (mine doesn't), the 100 ohm resistor should help to supress it and will also kill the potential.

good luck!

p.s.: one remark for those using [usually un-earthed] SAT-receivers as sources: the SAT-cable shield should be connected to house's potential compensation since otherwise these receivers usually have high potential especially in combination of a SAT-multiswitch.