838 Chapter 25
microphone amplifier far outweighs the hassle of a
similarly performing discrete transistor design, which in
this specific area is still its main close rival.
Every design case demands a long cool look to deter-
mine what sort of device makes most sense; there is no
one-fix cure-all technique, or device. For the most part,
the designs here are based on TLO-class or 5534-class
devices, determined mostly by whether low-noise,
high-output drive capability or high-input impedance
driving criterion. Modern devices (for a mature tech-
nology, new op-amps do seem to pop up almost weekly)
that fulfill these specific or other niche parameters
would of course be applicable.
25.7.3 Discrete Operational Amplifiers
The JE990, designed by Deane Jensen of Jensen Trans-
formers and manufactured by Hardy Co. of Evanston,
Illinois, is an example of an encapsulated discrete
amplifier module. Many fascinating solutions to op-amp
internal-design problems (some of which even IC
designers evidently haven’t realized existed) are imple-
mented in this design whose features demand a total
reappraisal of contemporary audio circuit design and
philosophy. Optimum input source impedance (normally
about 10 k: with most IC and discrete amplifiers) is
reduced to about 1 k: by the use of an IC multiparallel
input transistor differential pair. Small inductors in the
emitters provide isolation from potential high-frequency
instability due to the gain-bandwidth characteristic of
the first differential stage shifting with varying source
impedances. Unity-gain noise is a quoted staggeringly
low 133.7 dBu, while the output is capable of deliv-
ering full voltage swing into a 75: load. This permits
the use of exterior circuit elements of far lower imped-
ance, reducing thermal noise generation. This elegant
device inevitably carries a high price tag. Its many attri-
butes point to the direction for design. It is well ahead of
any devices available in IC form and also, to the
author’s knowledge, of any universal discrete circuitry
elements used to date in console manufacture. This
device begs the question of the wisdom of the complex
multiamplifier, multistage mixer configurations versus
true minimum-path circuit philosophy.
25.7.4 Instability
An unexpected thrill facing designers as they upgraded
to newer, much faster devices was the tendency for all
their previously designed circuits to erupt in masses of
low-level instabilities even in what had been perfectly
tame boards.
Layout anomalies, such as track proximity, were a
major contributor toward the stability problems, so new
layouts had to be generated with a whole new set of
conditions added to the already hazardous game of
analog card design. However, the real roots to this
problem are with the devices themselves and a lack of
appreciation of the relationship between their internal
configurations and the outside world. Everyone who
had been brought up designing around 741s had become
too used to treating them in a somewhat cavalier fashion
and for good reason. It was very hard work to make
them misbehave or even show a hint of oscillation.
People got used to treating ICs as plug-in blocks of gain
with little consideration for the fact that inside was a
real, live collection of electronic bits that still had all the
problems real electronics always had. The reason the
741 was relatively impervious to user-inflicted prob-
lems is analogous to the fact that it’s quite difficult to
get anything that is bound, gagged, and set in molasses
to not behave itself.
Mistake number one with the new devices was
believing that they were unity gain stable because the
data sheets said so. What that really means is “does not
burst into oscillation at unity gain (under these circum-
stances ...),” which is not the same thing at all.25.7.5 Phase MarginIt is important to maintain as large a margin as possible
between the internally structured gain-bandwidth
roll-off set for open loop and the roll-off around the
external circuitry determining the closed loop gain. This
is to preserve sufficient phase margin at all frequencies
for which the circuit has gain. Failure to do this can
result in the feedback being shifted in phase sufficiently
to become reverse phase to that intended (positive feed-
back) with oscillation resulting. Even if the phase isn’t
shifted quite that far, the feedback tends toward positive
and damped ringing when transients hit the circuit
ensues. Also, these resonance effects are extremely high
in frequency, typically many megahertz, so any radio
signal that gets as far as the circuitry will absolutely
adore an amplifier that is critically resonant at its
frequency! A reasonable phase margin to aim for at all
gain frequencies is better than 45°. In practice, a
compromise between desired circuit bandwidth traded
off against the need to tighten that bandwidth for the
sake of phase margins can be fairly easily reached with
the newer devices, provided the need to do so is recog-
nized.
There seem to be two schools of thought on band-
width versus stability phase margin. First there are the