AW: [sdiy] Seperate Grounds

[Czech Martin]

Ok, here comes my attempt to confuse you even more...

It has to do with mountains. You look into your "Tour de France book" and
read "Coll du Galibier 2732 m altitude" (hey, this is just an example, the
real number I haved forgotten).
What does that mean? 2732m above what? "Sea level" someone says from the
off. Ok, but what sea? In the British channel we have a tide of more then
12m up and down. Does this mean that the Galibier has a different altitude
in the morning then it was last evening? Ok, this example shows that people
need a reference level in order to talk meaningfull about altitude.
Before GPS there were different reference levels, say for Hamburg (Germany)
or Genua (Italy).

It's the same with electricity. "This battery has 2.7 Volts".
This statement is sloppy and it should be "this battery has a potential
difference of 2.7 Volts between it's clamps".
See?

Now, if you connect two circuits that exchange some
information via voltage levels, you have to tell them
what the reference level is. For example CD-Player and Amp.
You have two wires: a black one which you DEFINE as reference
and a red one you say it has the signal.

This is were my analogy ends. Because in wires, there is current flowing. No
information exchange without voltage change and thus current flow. The
current that flows toward the Amp in the red wire must flow back in the
black wire. Wires have resistance, so according to Mr. Ohm some voltage
difference (drop) will build up allong the wire as stated by U=I*R.
And of course you know that a wire loop has some inductance, too. And it has
some capcitance between wires and maybe to other objects near by. These
demand that Ul=L*d/dt(I)
and Uc=1/C * INT{I}dt.  Voltage across an inductor is inductance L times the
change of current in time. Voltage across a capacitance is the integral of
current over time (the charge) divided by capacitance C.

But so far, there is nothing wrong, because our voltage drops
on the black reference line (as well as on the red line) have to do with the
signal, the signal may get weaker, the signal may have some roll off in the
highest frequencys, but usally no real problem with -say- 30 cm length.

The trouble starts, when people try to save money.
I.e. they use the black wire (which we defined as reference)
as return path for other currents, too. E.g. for lamps, relais
etc. This is a pretty bad idea, because now the three possible
voltage drops have nothing to do with our signal, they are percieved as
noise. Clicks, bleeps and the like.

So, if I have a pot connected to some voltage and (hmm) reference, and feed
that into an amp, I should make sure that the pot reference level is that of
the amp (local), and that the pot voltage has also to do with the reference
level at the amp. Otherwise the voltage difference between pot wiper
and amp reference will not be the same as measured across the 
pot wiper and pot reference. Some times you know that there is 
some difference induced by your wiring scheme, but it is perhaps constant,
due to constant currents causing ohmic drop.
You could then say that this will not be harmfull.
But AC wiggling will be sure a problem for you and your pot.

There is not the "correct" solution, you have to draw up a schematic with
all interconnects modeled as resistors and perhaps inductors, you may decide
to leave away the wire capacitance. Then observe where the currents flow, if
they are AC or DC and the magnitude. If you draw a schematic with
interconnects as ideal wires (the black lines typically used in a schemo) it
is of course impossible to understand the problem.

Another problem is long lines. Now magnetic and electric fields can get into
your line via capacitive or inductive coupling. Today, our environment is
filled with electro-magnetic energy of all kinds. So the meaning of "long"
depends on your environment. And also to the grade off acuracy you need.
Today CD recording is common even for amateurs. That means 16bit or 96 dB
S/N. That IS demanding.
You can use a shielded wire (coaxial) to avoid capacitive coupling. This is
what "HiFi" gear does. Unfortunetaly the reference wire is now the outer
shielding skin. This means that
the capacitive coupled shield currents will now flow in your reference line,
thus causing voltage drops, thus noise.
The reason why this configuration works most of the time is,
that the Hifi gear and the noise source fight against on the shield, a
capacitive divider. Usually the gear capacitance to
earth ground will be larger then the noise source coupling capacitance into
our shield, so the noise level the amp sees may well be below -100dB.
Magnetic coupling is especially nasty, you can't shield it.
Well, you can, at high frequnecies. Because there the magnetic field is
strongly connected with the electric field. A thin sheet of alu or copper
will have eddy current if a magnetic hf field trys to permeate, these
currents have in turn a magnetic field in the oposite direction which will
weaken the incoming field very soon. 0.1mm of copper is enough.
Unfortunately there are notorious magnetic noise sources near DC,
transformers, electric engines etc. 50/60Hz. There the eddy currents are
weak, you would depend on magnetic flux "bending" and need very very thick
high permeable shields. Have you seen the shielding for magnetic resonance
radiology chambers? 30cm and more. We can't do that. There is only one thing
that helps:
distance. Put transformers etc. far away. If a long line has problems with a
magnetic noise source and if no change of the geometry is possible, we can
play a trick on the evil magnetic field: we use a twisted pair wiring.
Because now the field will come into the wire loop sometimes in this
orientation, sometime the other. If we make a good job (perfect symmetry of
the twisted pair) the differnt induced components will cancel.

Another problem is ground loops. This is the last problem and really a nasty
one. Consider our HiFi gear, CD and Amp.
Now , the amp usually has a metal chassis, we assume that it may be
connected to protective earth, which is basically a good thing. Usually the
manufacturer will separate chassis protective earth from signal ground. We
assume, that this is not the case, perhaps it is an iron less tube amp for
electrostatic speakers. We do not want that the black wire
may rise to 500 V plate voltage due to fault, so we connect
it to chassis protective earth. Connecting the chassis is also
good, because it is no longer floating. {Sensible persons can feel electric
currents on a metal, floating chassis (capacitive voltage divider earth, hot
supply, chassis, person, earth).
I have an IKEA lamp at home, the stand is floating metal. If I touch it very
soft, I can feel the current in my fingers. A stronger grip will shunt it to
ground vias my body resistance.}
Only a grounded chassis will make it possible to build in a line filter
against unsymetric AND symetric power line noise.
Of course, this costs a few bucks and is therefore omitted.

Ok, back to example. So far there's nothing wrong. It is only that the black
line is now conected to protective earth at ONE point in the amp. In the
other examples it was floating in respect to earth. This will not change the
voltage difference between red and black wire. If the black wire is formed
as coax shield, this may be beneficial, since capacitive coupling will not
modulate the shield voltage against earth. OTOH now capaictive coupling to
the shield may lead to larger shield currents, since the shield is signal
reference this can cause noise volatge drops. We assume that this is no
problem.
Now assume that we exchange our CD for our beloved vinyl player. There's a
motor connected with mains in it, so the manufacturer made a metal chassis
connected to protective earth. And he choose for some reason to connect the
signal ground to chassis as well. For some reason we plug in the player into
another outlet, that has a complete different wiring branch then the amp
has. Voila: a perfect ground loop.
It will not really humm with 50/60 Hz, it will buzz.
What the heck is going on here? One end of the black wire is connected to
the phono protective earth. The other is connected to tube amp protective
earth. Guess what? Who says that these two protective earth points have no
voltage across them, you bet they have. There's all kind of return current
flowing in earth, it is 50/60 Hz and also a considerable amount of harmonics
and simply higher frequnecy noise. This voltage across our black wire is our
noise. The black wire can't really short it. Protective earth has massive
wire gauge, for a good reason. Therefore it makes Buzzzzz. What can we do
now?
Some people decide not to live too long, in order to give the social
security system some relief, so they tape the protedctive plug on one end or
the other or both. In case of a fault your equipment will get excitingly
mmmhh -say - hot. The sensible people use a transformer. One audio
tramsformer is enough. We put it near the amp, it doesn;'t really matter. It
will convert current into usefull magnetic flux and back into output voltage
again.
But: our ground wire is now split. The player black wire gets into one
transformer coil and back as player red wire.
The same on the other side. The wires on the different sides of the
transformer have NO metal path against each other.
Sure, the player will wiggle up and down relative to the amp,
since both reference levels (protective ground) are not the same, but since
red and black wire wiggle in the same fashion this will lead to no
additional transformer coil current.
The transformer has the advantage that the potential difference  across both
sides can be very large, some allow for 1000V or more. This is foolproof.
OTOH transformers have problems with low frequency signals, they cause
distortion and also some noise due to core losses (anything that has losses
will cause thermal noise). And good transformers are expensive.

Now, if you need equipment that always works, in the mud and rain, open air,
whatever, you do the following:

transformer , one on each side. So even large voltage differences will not
be harmfull. Open air, wires can be 300m , 500m and longer...
Use twisted pair wiring. Thus magnetic fields will cancel.
Use a seperate coax shield. This is connected on both sides to the metal
chassis. Any capacitive currents will be shortet.
The shield must not be floating in order to keep electric fields out. The
shield is NOT involved with signal transportation.

Notice something? The two signal wires are now completely symmetric. We say
balanced. Which is signal, which is reference? You can't tell. In between
the transfomers they
are floating in respect to earth. Voltage is potential differnece, in this
case between these two wires. The balanced scheme ensures that the twisted
pair principle is 100%,
i.e. all magnectic and possibly rests of electric fields
will cancel in their effects or only lead to common mode
noise, which is rejected by good transformers. Symmetry is the reason you
need two transformers. They should have also a stray shield
winding foil between the coils, thus preventing any capacitive
speak through of common mode.

Since good transformers ar bulky, expensive and still have signal problems,
the common mode suppression action is often mimiked by active electronic
circuits, balanced line drivers and difference instrumentation amps. You
know, this XLR
stuff in studio gear. But: these electronically balanced
things will not tolerate large potential differnces across as transformers
do. Inside your studio rack ok, with 1000m cable
across a NO NO.

Still with me?

You need to understand Ohm's law, and capacitor and inductor (transformer)
action. You need to know what you are doing. There is no definite answer for
all situations.

You need to draw a "real" schematic that explicitly depicts all wires as
resistor, capacitor and inductance and sometimes as transformer network
(this transformer has nothing to do with noise rejection, but is symbolizes
magnetic noise coupling). You need to draw possible noise sources. Make the
schematic as easy as possible and as complicvated as necessary. If magnetic
flux is absolutely unimportant, leave away all coils and transformers etc.
If it is all DC , leave away the caps. Then you will see what's going on and
what the optimal wiring scheme for your application is, regarding cost, S/N
performance, geometry and so on. For different people, demands and
applications the optimal solution can be different.

That's why banana, shielded 6.3mm and balanced XLR people will never agree,
for example.

Ok, back to the Tour de France, now, the riders should be just at the
start...

m.c.