912 Chapter 25
device available, so the primary noise modes are those
above-mentioned device vices. This certainly isn’t too
difficult with FET front-end devices, with their high
OSI (optimum source impedance). These devices have a
couple of other major benefits in this application though
by virtue of their FET inputs. Input current (hence,
input current noise) is extremely low, and being FETs
they don’t have the many low-frequency junction and
surface noises inherent to bipolar devices. It seems a
paradoxic absurdity to use an ultrahigh input impedance
device for zero impedance mixing, but in many ways
and under some circumstances they’re better suited than
bipolars. On the other hand, the intrinsically superior
noise performance of a 5534-class device can pay divi-
dends in this application. Like so many cases in console
design each individual application needs staring at for
its own optimum solution. This is all really only a
problem for those who have the luxury of designing
small mixers or where it is more or less guaranteed that
only a small number of sources will be allowed to hit
the bus simultaneously and hence where the parallel
impedance of the sources remains fairly high. In most
midsize and large consoles without these constraints,
mix device noise will likely predominate. Device choice
will be down to its self-noise (of course), and output
current capability if the summing resistor value is low,
and ability to cope with a big hairy capacitative bus
sitting on its input current node. Integrated mic-amps
have been successfully used as differential passive
mix-bus amplifiers, which with their very low OSIs
stand a chance of getting closer to that low bus imped-
ance and low bus noise nirvana. However, as alluded to
earlier, the channel-off noise contribution from all those
bus-driving amplifiers in all those channels is more
likely to then predominate. It is a balancing act.
If bus noise performance truly is a major concern (as
it could possibly be on a tracking console) removal—as
in physical disconnection—of all unused sources from
the bus at all times is the best way to get the noise gain
down and that bus impedance back up to where
mix-amp noise can be optimized to it. No way to run a
railroad or a mixdown console, though.
Things can get a bit startling if the resistance/OSI
relationship is awry. Above the OSI as much as below
its OSI, device noise becomes an increasingly important
noise contribution. Many years ago in a mixer design
with bipolar device mix-amps and quite high mix resis-
tors, the measured bus noise was actually quieter on a
20-channel version than on the 10-channel original. It
wasn’t until much later that what was actually
happening finally dawned. Increasing the number of
source resistors reduced the bus impedance, previously
well above the OSI of the amplifier with only 10
sources, to closer to the OSI, where input noise voltage
was contributing less.
Theoretical source impedance and device contribu-
tion tell less than half the story in a practical design.
They may be quantifiable in the isolation of a test
bench, but thrown into a system they can all seem a bit
meaningless. It’s all largely a matter of grounding and
out-of-band considerations.25.13.3 Radio-Frequency InductorsInductors are used between the bus and the amplifier
input in Figs. 25-86 and 25-87. A simplistic view is that
they are there to stop any radio frequency on the mix
bus from finding its way into the electronics, but this is
only part of their purpose. The ferrite beads and small
chokes (about 5μH) are there to increase the input
impedance and hopefully help decouple the bus from
the amplifier at very high frequencies. The larger induc-
tance creates a rising reactance to counteract the falling
reactance of the bus capacitance. If left completely
unchecked, this capacitance would cause the mix-amp
extreme high-frequency loop gain to turn it into an RF
oscillator. Feedback phase leading around the amplifier
stops the gain from rising, but if it were not for some
series loss (accidental or deliberate) in the input leg, it
would be insufficient to hold the phase margin of the
amplifiers within their limits of stability, especially at
bandwidth extremes where device propagation delay
becomes significant in the loop. A small series resis-
tance can provide this loss while also defining the
maximum gain to which the circuit can rise. A parallel
inductor-resistor combination improves on this in a few
important respects.
The inductor is calculated to present low in-band
(<20 kHz) reactance, allowing the mix-amp to operate
on the bus in its intended virtual-earth (zero-impedance)
configuration. The reactance rises gently at the audio
high-frequency end, imparting little frequency response
anomaly but a definitely beneficial partial phase
straightening against the inevitable effect of heavy bus
capacitance.
At even higher frequencies, the inductive reactance
continues to rise until the combined network imped-
ance is limited by the resistor, which is of high enough
value to define amplifier out-of-band gain to a reason-
ably low value. It is low enough, however, to stop the
inevitable inductor-bus capacitance resonance from
getting completely out of hand. Making a stable induc-
tance-capacitance oscillator is one way of preventing