IT’S WELL KNOWN THAT audio power
amplifiers like to get a good set of
grounds, or noise around the inputs
may not be rejected properly, causing
hum and buzz. So when a guy called
me asking how to clean up his interface
from his clean audio signals to
his LM3886 power amplifier, whose
ground system was pretty noisy and
lumpy, I thought for a second and
replied that the solution was easy
(Fig. 1)!
The set of four 4.02-k resistors
(1% R’s but matched to 0.1%)
make up an adder-subtractor
circuit—straight from Philbrick’s
1955 Applications Manual. Let’s
stand at the power ground for our
scope ground. If there is a quiet
signal “out,” referred to its analog
ground “A,” fine. But we must
beware of any common-mode (CM)
noise, which would appear between
the A ground and the PWR ground.
This noise might be related to any
kind of ambient noise, or noise on
the power amplifier’s supplies, or
(in a vehicle) automotive transients
and radio-frequency interference
(RFI). I told this guy, “This circuit
will reject any and all such noises.”
The old equations give you:
Amplifier VIN – (VPR) =
V (signal out) – VA ground
I recommended an LME49710,
which has 50 MHz of bandwidth
and good clean gain and linearity
and noise rejection, per www.national.com/an/AN/AN-1671.pdf.
I recommended a “gimmick” of
teflon twisted pair as CF, starting at
6 in. and unwinding it as needed. I
mean, we don’t know exactly what
kind of noises there will be, or how
much of the wiring strays. Still, the
op amp should be supplied with
reasonably quiet power that is tied
to the PWR ground. The 4k resistors
do not have to be closely matched
to ensure gain accuracy—but to
help give you good common-mode
rejection ratio (CMRR), much better
than 1% R’s will do. If you wanted
80 dB, you’d have to trim them.
OPPORTUNITY
After I hung up, I could not
stop thinking of this “problem = ~
opportunity.” Why does this look
familiar? I thought about some of
the problems my colleague Nick
Gray had been trying to solve over
the years. Hey! this looks just like
the problem with analog signals
that need to be sent to an analogto-
digital converter (ADC)! You have
an analog ground plane and a digital
ground plane. But if you try to
just strap the grounds together,
you’ll get absurd noises.
And if you have one ADC, the
CM noise rejection can be bad.
But if you have one analog ground
plane and one big digital ground
plane that are serving two or four
or more ADCs, the CM noise coupling
can be horrible! What’s a
mother to do?
The adder-subtractor shown in
Figure 2 will reject the CM noises,
neatly, using one adder-subtractor
per ADC channel. The clean, quiet
voltage that is sent in between the
signal and VA will re-appear at the
adder-subtractor’s output, referred
to power ground, for each channel.
The old equations give you:
VOUT – (VDIGITAL GND) =
V (signal out) – VA ground
Now, the LME49710 mentioned above may have a signal
bandwidth of just 20 MHz,
plenty for audio or for
some ADCs. It may be trimmable
to a CMRR of 70 dB
out to 2 MHz. That’s what I
saw when I actually built it.
But what if you need a fast
ADC and a lot of CMRR versus
frequency?
A KEY MODIFICATION
Let’s swap in the
LME49713, a fast currentfeedback
amplifier (CFA),
which will pass signals up
above 90 MHz, and we
may be able to get decent
CMRR and noise rejection
out past 20 MHz. (The
resistors have been cut to
2k to make sure you can
get full bandwidth. It might
run even a bit faster if you
chose 1.2k.)
The LMH6714 can go
out to even faster, 400
MHz. But the ’49713 may
have better linearity. Hard
to tell. When you need
to do precision work plus
fast bandwidth, everything
has to be engineered and
tested. The circuit might
change slightly if you need
fast, clean step response
rather than just wide bandwidth
for sines.
Will that be fast enough
for a pretty broad-band
ADC? Well, that will depend
on the actual circumstances
of your system needs.
I mean, you could always
have more CM noise than
this adder/subtractor could
quash. I did my testing
with 10 V p-p. Could this
be used in addition to a
Balun? Insert that ahead of
the first two 2k resistors.
Probably. I mean, when
things get fast, then you
always have to be prepared
to do some real engineering.
Still, this is one good
tool to add to a good
toolbox, when you have to
accommodate (and reject)
many kinds of nasty noises,
conducted and induced
from radiated noise.
For this case, adding a
few pF of Cf seemed to be
doing more harm than good,
so I installed several inches
of “gimmick” = twisted pair
as a capacitive CMRR trim,
for the signal path going to
the positive input.
What’s the big deal with
the “adder-subtractor”?
When George Philbrick
developed the K2-W, which
was one of the first operational
amplifiers with differential
inputs, it facilitated
simple adder-subtractors
that did not need a dozen
resistors and two or more
chopper-stabilized amplifiers
and hundreds of watts.
George never had access
to any 400-MHz op amps
as we can easily buy these
days. He’d be impressed
with modern op amps—and
applications circuits.
Can these amplifiers
provide a voltage gain other
than 1.0? Sure—and you
have to engineer it. And
to get good results, you
always have to plan a good
layout. This is just a start,
to indicate all the things
you can do.