[John Byrns]

I remember the guy that originally posted that article, he used to post
on the usenet newsgroup "rec.audio.high-end", it seems unlikely you
would have found it on the web.


Regards,

John Byrns


In article ,
Henry Pasternack wrote:

This is something I found on the Web. It was apparently posted on the
usenet, perhaps to this group or maybe rec.audio.tech, many years ago,
althought I'm unable to find any evidence of the article in the Google
archives.

I offer the article without comment, and don't make any claims as to
its accuracy. Perhaps there is some value to what is written, as quite
a few people have linked to or reposted copies on their websites. The
author has left his phone number and email address in the signature. If
you have any issues, you may wish to contact him there. I would, in fact,
be very interested in hearing if anyone manages to contact him that way.

-Henry

--- BEGIN REPOSTED ARTICLE ---

Someone suggested I write an article on power and grounding techniques
for equipment construction. This is a topic I can't really do justice to;
whole books have been written on the subject. In the spirit of reckless
misinformation, I'll try to summarize a few hints. Consider this an
introduction for advanced beginners, but bear in mind there is room for
expanding and improving the information contained below.

-Henry

* * * * * * *

If circuits worked exactly as they are drawn on schematics, there'd
be no need to worry about power and ground topologies or wire routing.
Instead, we could just wire up all the components willy-nilly making
sure only that all the specified connections were complete. In the
real world of electromagnetics, things aren't quite so convenient:

1) Ground isn't really ground. Ideally, ground is a zero-voltage
reference that never varies. In fact, every ground bus that
carries current has small voltage variations across its length
due to small but significant resistance.

2) Stray capacitance. Two parallel conductors form a capacitor
across which unexpected AC currents can flow.

3) Electromagnetic induction. By definition, every circuit forms
a loop. When a changing magnetic flux (due, perhaps, to the
strong field from a transformer or choke) exists in the loop,
a current will be induced.

Now I'll talk briefly about the consequences of these real-world effects.

Ground Loops.

A ground loop occurs when a ground circuit encloses a loop and a
source of energy causes current to flow through the loop. If the flow
of current causes a voltage drop in a sensitive gain stage, the
voltage can be amplified and appear as noise. A typical ground loop
occurs in the amplifiers I just built. Because I use grounded power
cords, the chassis of my monoblocks are joined electrically at the
power outlet. The signal grounds are connected (via the input jacks
and interconnect cable shields) at the preamp output terminals. The
signal grounds are also connected to chassis ground internally in each
amplifier. The result is a loop that picks up inductive noise or
leakage currents and couples them into the amplifier gain circuitry,
causing an audible hum in the speakers. The quick solution is to
break the loop by using a three-to-two wire converter on one power
cord. A better fix is to "lift" the internal connection between
signal and chassis ground in one or both amplifiers and install a 10
Ohm resistor. The resistor is a significant obstacle to the flow of
current compared to the stout power cords and effectively breaks the
loop.

Another sort of ground loop occurs internal to equipment, but for
a different reason. In any circuit, one can trace the flow of power
from the supply, out to various circuits, and back to ground. Each
power supply typically provides current to more than one circuit
branches. Topologically speaking, the branches form intermeshing loops
that originate and terminate at the power supply. In the regions
where the loops overlap (share common conductors), voltage drops in
one loop can impose unwanted signals in another.

When a circuit loop overlaps on the supply side of a power source,
the designer will use parallel capacitors and/or series resistors or
inductors to block or shunt away signal currents on the supply rails.
The series components keep unwanted signals from leaving their circuit
branches, and the parallel caps redirect those that escape to ground.
Purifying the supply rails in this manner is called "decoupling".
Decoupling also helps to rid the rails of noise that creeps in due
to magnetic or electrostatic fields (more on theis later).

On the ground side, it's very important to keep the return currents
separate from one another as they work their way back to the common
power supply terminal. Practically speaking, we want to minimize
overlapping loops in the ground circuit. This is done by using a
short, thick ground bus of very low resistance and connecting all
ground wires to it. The most effective solution is to provide
separate conductors for each ground current source and return them all
to a single point; this is known as "star" grounding and is standard
practice in audio design. In tube gear wired point-to-point, it may
be sufficient to route a heavy ground wire around the edge of the
circuit and make all connections directly to it at the closest
convenient point.

Inductive and Capacitive Coupling.

Every conductor carrying a signal current generates an electromag-
netic field that can couple to other circuit elements. In addition,
the environment in which equipment is used is loaded with power line
and radio frequency fields that can get into sensitive circuits and
cause noise and distortion. I'll give some very quick information on
minimizing these problems.

Electrostatic fields and capacitive coupling can be reduced by
distance and shielding. A grounded metal shield stops a static field
in its tracks by imposing an equipotential barrier. To keep the field
from sneaking past the shield, the sensitive circuit must be
completely enclosed. Capacitance decreases as the reciprocal of the
distance between conductors, so separating sensitive wires as much as
possible is the other solution to stray electrostatic coupling.
Keeping circuit impedances low will help a great deal by raising the
frequency at which the coupling becomes a problem (hopefully beyond
the bandwidth of the circuit or offending signal source).

Shielding with non-ferrous metals is useless against magnetic
induction. The answer to induced hum of magnetic origin is to keep
circuit loop area to a minimum. The smaller the loop, the smaller the
volume of magnetic flux enclosed and the lower the induced current.
For this reason, power and signal wires should be twisted tightly with
their ground conductors whenever possible. This increases capacitance
between the conductors, so watch out in high-impedance signal
circuits. Plan the layout of hum-sensitive circuits so that the
enclosed area is kept to a minimum. Place transformers, chokes, and
high-current conductors as far as possible from these circuits. Iron
or mu-metal shields may be of some help, but I wouldn't count on them.
It may be better to put the power supply in a separate box for
sensitive preamps.

In most cases it is good advice to keep wire runs short, to twist
conductors carrying opposite currents together (signal and ground,
power supply lines, filament wires, etc), to route wires near the
chassis, and to physically separate high-gain stages from components
and wires that carry high currents.

Practical Advice.

When laying out your circuit, look at the schematic and identify
all of the separate current loops from supply to circuit to ground.
Then come up with a scheme to keep these loops as separate as possible
on the ground side. The biggest source of ground current in most
amplifiers is the power supply ripple in the main filter capacitor.
This current, which contains noisy, high-current rectifier current
spikes, usually flows back to the power transformer through the
secondary center tap. A good strategy is to connect this lead
directly to the filter capacitor common terminal. Then run a short
stub (could be a half inch) to a second point that will serve as a
star ground for the signal circuitry. Another source of high current
ground flow in a power amplifier is the common lead from the speaker
terminals. You'll want to make sure the low-level signal star ground
does not have output ground currents flowing through it or you'll
increase the possibility of instability. What I'm describing here
is really a "star of stars" grounding scheme. If you view every
conductor as a resistor and try to keep associated voltage drops
from crossing loops, you'll end up with the best ground scheme.

Here's the grounding topology in my tube power amps:

The main electrolytic filter caps and driver decoupling caps are
all mounted on one circuit board. The three decoupling caps share
a common, wide circuit board ground trace. The ground for the main
filter cap, for the transformer center tap, and for the decoupling
caps all come together in one small area of the circuit board. The
driver stage ground points are located close together near the central
ground, but "out of the way" of the rectifier current flow. I slotted
the copper strategically to "steer" the charging current current away
from the driver-stage grounds (defining the stub conductor described
above).

The following wires attach to the driver-stage ground area on
the filter cap board:

1) A single ground wire from a local "star ground" on the driver
board that connects the input jack shield, global feedback
cable shield, and the input stage bias resistors grounds.

2) A single ground wire carrying the return current from the
B- supply. This supply is floating and connected to the
driver board by a tightly twisted pair. B- is bypassed
locally by electrolytic and film capacitors. A second
"star ground" on the driver board connects all the bypass
caps. The rationale is that low frequency hum currents
are handled by the electrolytics and don't make it off
the filter circuit board. Local bypass caps keep RF
signals on the supply lines from getting onto the driver
board. The current-sourced differential design greatly
reduces the coupling of the input and driver signal currents,
in the ground leads, justifying the use of a single B- supply
and ground wire. I agonized over this layout for a while,
but it seems to work quite well.

3) A ground wire from the negative bias board. The bias supply
is also floating and connected to the bias board (which also
holds the output coupling capacitors) via a twisted pair.
There is a 0.47uF film capacitor bypass on the bias board.

The common lead from the output tube cathodes connects to a
ground point that runs directly to the main star ground without
passing through the driver ground node. The power transformer
center tap comes in on its own stub directly to the main capacitor
common terminal.

The two B+ wires to the driver stage are very short and connect
directly to the driver board. The output stage B+ is supplied to
the output transformer primary center tap from a circuit area close
to the main filter cap positive terminal. The power cord ground
wire is connected to the chassis at a convenient lug near the back
panel AC socket. The filter cap board has a ground wire to a lug
on the chassis as well (and perhaps this is why I have a ground loop;
I need to experiment with the location of my connection from signal
ground to chassis ground).

A note on solid-state Class AB amplifiers and power rail routing.

The current drawn by the two halves of a Class AB output stage
is rectified and has heavy harmonic components. Because solid-state
amps draw a lot of current from the rails, there is a good chance of
inducing signal-related noise in low-level circuitry unless careful
layout is used. A good technique is to twist the positive and
negative rail leads together and route them directly to the output
devices, keeping them away from low-level conductors. The rails are
usually bypassed on the circuit board by electrolytic capacitors
which should have a dedicated ground wire for their common terminals.
This wire should be kept apart from the small signal ground lead.
Care should be taken to keep the negative feedback connection, which
typically connects to a node near the output output transistors, from
coupling to the power leads. A sophisticated design will use multiple
electrolytic capacitors mounted right at the output devices and
clever circuit board layout. I believe the famous Analog Devices
application note on a wideband current feedback amplifier has some
useful notes on this problem (does anyone have a reference?).

--

Henry A. Pasternack
Member Scientific Staff (514) 761-8734 (phone)
Bell Northern Research, Montreal (514) 761-8509 (fax)

--
Surf my web pages at, http://fmamradios.com/