Page 26 - Analog Circuit Design Art, Science, and Personalities

George A. Philbrick

poses were served by these structures, just as with simulators today. The next stage,
in which the real control mechanisms were replaced by models, permitted the whole
loop to be electronic and hence vastly more flexible and greatly accelerated. In
simulators of this sort, several plants might be interconnected under control, so that
the newer stability problems thus encountered could be studied conveniently. Again,
plants with mu1 tiple inputs and outputs having internally interacting paths were
included, and regulatory loops in hierarchies where master controls manipulated the
desired conditions of subordinate controls, all could be simulated in an analog. Note
the ascending succession of feedback loops, which are most dramatically repre-
sented in control systems of this sort: within amplifiers to attain promptness and
stability; locally around amplifiers to give the desired mathematical performance
for regulatory mechanisms; in control loops to promote the minimum difference
between desired and existing conditions; in more comprehensive control loops
which include complete but subordinate loops in cascade; in still more comprehen-
sive loops for supervisory or evaluative purposes; and finally in the experimental
design and optimizing operations, using models or computational structures to
evolve most effective system operation.
Servomechanisms are also part of the lore which preceded and inspired the
modern analog machines. Though not as old as the governors, pressure regulators,
and controllers of temperature, flow, level, etcetera of the last paragraph, servos as
positional followers were functionally similar as regards control philosophy and
feedback loops. Further, being more modem, they benefited from the increasingly
mathematical technologies of development and design. Perhaps most relevant was
the simultaneity and parallelism between servo theory and that of feedback ampli-
fiers in communications. Stability criteria for the latter were seen as applicable to
the former, at least in the linear realm. Analysis in the frequency domain, a natural
procedure for linear communications equipment, was carried over rather directly to
servomechanisms. This debt has since been partially repaid. as servomechanisms
have helped to furnish nonlinear analog elements and other items in computing
equipment for the study of nonlinear phenomena, generally in the time domain: as
they occur in communications and elsewhere. Thus do the various doctrines and
practical disciplines feed on each other to mutual benefit, and (if you will forgive
the Liberty) feedback sideways as well as back and forth.
We pick up servomechanisms again, much further back along the trail, and usu-
ally in relatively low-performance embodiments. Though scientific instruments do
practically everything today, including computation, synthesis, manipulation, and
regulation: on every scale, they were once used principally for measurement, in the
laboratory or the observatory. For accurate measurement it was found that feedback
methods, when possible, were surpassingly effective. While the underlying philo-
sophical reasons for this circumstance are of vital importance, we shall take them
here on faith. Note. however, that the observation of balance in a measurement, and
the manipulation which may be made to achieve balance, is still a feedback process
even if done by a human agency. The slave can be the experimenter himself. Precise
weighing with a beam balance may stand as a clear example of this procedure, but a
myriad of others may readily be spread forth. Succinctly, the process is reduced by
feedback to dependency on only one or a few reliable elements. Automation of the
loop-closing, null-seeking action merely replaces one slave by another. In this light
the venera.ble self-balancing slidewire potentiometer recorder stands with the latest
feedback operational amplifier, and so we see yet another plausible path from then
to now.
Antedating but partly anticipating the development of active analogs was the use

7