842 Chapter 25
standard voltage-follower configuration and that this is
the most critical configuration for stability, it is not a
preferred circuit element. The manufacturer will have
designed the IC to be just stable enough at unity gain to
be able to say so unblushingly, but with probably little
real-world margin to spare. Hanging a compensation
capacitor across the appropriate pins will slow up the
slew rate and not necessarily make the whole amplifier
any less unstable. It is better not to tempt fate.
25.7.10 Input Saturation
The use of a standard voltage follower implies that in
order to maintain the same system head room in that
stage, the input has to rise and fall to the same potentials
that the output is expected to. It can’t. In most op-amps,
especially those with bipolar inputs, the differential
input stages saturate or bottom significantly before the
power supply rails are reached and certainly before the
output swing capability is attained. This limited input
common-mode range means that the follower not only
will cease to follow but will also spend a considerable
amount of time in unlatching from one swing extreme
or the other. Once an amplifier internal stage has
latched, the feedback loop is broken; the stage has no
assistance from the servomechanism to unstick itself.
Once the loop is reestablished, it has to settle again as if
from a hefty transient before it can resume following.
Basically, this is an ugly scene. Uglier yet is the propen-
sity of some devices when the input common-mode
range has bottomed for the output to lunge to the oppo-
site rail. Talk about “sonic character.”
IC manufacturers commonly specify the
common-mode input voltage range and it is precisely
this limit that would be exceeded in use as a follower.
For reference they are: ±13 V for the 5534, ±11.5 V for
an LM318, and +15 V to 12 V for a typical BiFET.
All fall far short of the power supply maxima.
Provided enough gain is built around the amplifier to
prevent these common-mode limits from being reached,
there should be no latching hangups; the feedback
network also provides some substance to hang
closed-loop compensation around in addition to
enabling the full output voltage swing of the amplifier
to be utilized.
Similar settling-time problems occur any time any
stage is driven into clipping, but given the high
power-supply voltages and consequent large head-room
common today, clipping should be rare.
In short, not only for this good reason, the standard
voltage-follower configuration is pretty bad news.
25.7.11 Front-End InstabilityAltogether the most obscure potential insta-
bility-causing effect relates directly to the behavior of
the input stage in bipolar front-end op-amps. The
gain-bandwidth characteristic of the input differential
stage is greatly dependent on the impedance presented
to the input, the gain-bandwidth increasing with
reducing source impedance. There is the possibility that
given an already critical circumstance, the erosion in
phase margin due to this effect can cause overall insta-
bility. This instability can be mitigated by limiting the
gain-bandwidth excursion by means of a resistor (typi-
cally 1 k:) and/or some inductance in series with the
input. Ordinarily, this would have little effect on circuit
performance but may, especially in microphone ampli-
fiers, detract from noise performance. Noise perfor-
mance is largely dependent on the amplifier being fed
from a specific source impedance, and 1 k: would be a
sizable proportion. However, it’s usually fairly easy to
arrange in the design stage such that the IC doesn’t have
a zero impedance at either of its inputs.
Fortunately, because of the far greater isolation
between the FET gates and their channels, this is a
problem that FET-input op-amps do not have. A similar
approach to that proposed for output isolation (i.e., an
inductor rather than a resistor) in series with the affected
input seems, on the surface, an equally good idea. The
impedance of the inductors would be low at audio
frequencies (so not affecting noise criteria signifi-
cantly) and high at radio frequencies where the low
source impedance phenomenon does its work. Unless
the value is critically defined, an inductor of sufficient
value to provide a usefully high reactance at RF also
could be self-resonant with circuit stray and its own
winding capacitances at a frequency probably still
within the gain-bandwidth capability of the amplifier.
Takes a bit of care.
Those who have experienced design with discrete
circuitry will not be surprised that this source imped-
ance instability effect is also the reason emitter
followers are the most instability prone of the three
basic transistor amplifier configurations. The cure is the
same. Not only does the series impedance limit the
source impedance before zero, it also acts together with
any pinout and base-emitter capacitance as a low-pass
filter helping to negate further external phase shift that
may detract from stability. This base source-impedance
instability is quite insidious in that it can either
contribute to instability of the amplifier loop if it is
already critical or it can be a totally independent insta-
bility local to the affected devices with nothing whatso-
ever to do with the characteristics of the external loop.