[sdiy] Matrix 1000 grounding question

[Tom May]

So I've got this Matrix 1000, and they seem to be noted for their hum.
As it turns out, the hum doesn't vary with the volume setting which
makes it look like a grounding problem.

Indeed, here is the output section of the M-1000.  I'm not really sure
about the size of the inductor but I don't think it's important for
this discussion:

             +
TL081 out ----||-----/\/\/\/--+---@@@@@@---+--------------------- output
             4.7uF   100ohm   |    20mH?   |         +-------+--- jack
                              |            |         |       |
                         1nF ===      1nF ===      signal   === 1nF
                              |            |        gnd      |
                            earth        earth             earth
                             gnd          gnd               gnd

The 1nF cap at the output seems to be the only connection between
earth and signal grounds.  Earth ground is the chassis, connected to
the green wire in the plug.

It turns out that if I use a 3-to-2 adapter on the Matrix's power
cord, the hum goes away completely.  But I would prefer to leave the
safety ground in place.

So I tried connecting the three capacitors to signal ground instead of
to earth ground.  This also solved the hum problem, but left the
signal and earth grounds with no connection at all which doesn't seem
like a good idea.  This is where my understanding gets fuzzy.  I
believe the reason for connecting the grounds with various
combinations of caps, diodes, and maybe even resistors have to do (at
least) with using the chassis as an RF shield, and with possible
faults between the primary and secondary windings of the power
transformer.

It turns out that connecting the two caps in the pi filter to signal
ground and connecting the cap at the jack to earth ground also solves
the problem.  In retrospect, this is perhaps obvious, since any signal
on the chassis will be coupled into the output via the caps (right?),
but I guess I was giving the designers the benefit of the doubt.

Does this seem like a reasonable way to have things grounded?  How
does the capacitor allow the chassis to function as an RF shield -- I
don't see the complete circuit that allows current to flow in the cap?
Maybe this some RF and/or ground loop voodoo.  If this were running
off a wall wart would I just want a wire between signal ground and the
(non-earthed) chassis, or nothing at all?  And what were the designers
thinking when they connected the pi filter caps to earth ground?

Tom.