Page 25 - Analog Circuit Design Art, Science, and Personalities
P. 25
Analogs Yesterday, Today, and Tomorrow
lapped to breathe life into this department. Among others were Physics and
Scientific Instruments, Communications and Electronics, Controls and Servo-
mechanisms, Mathematics, and Aeronautical plus Electrical plus Mechanical
Engineering. It is recognized that these fields are not mutually exclusive, and that
each realm constitutes a multidimensional cross-section which has interpenetrated
the other realms enumerated.
There is one thread, come to think of it, which appears to run through the whole
background of the analog doctrine, and which may be said to belong to it more
intrinsically that it does to the other major branch of computation: that thread is
feedback. It will appear again frequently in what follows.
The clearest anticipation of analog machines was in the differential analyzer.
This primarily mechanical device could handle total differential equations at least
as well as we can now, and in some ways better. One such analyzer afforded auto-
matic establishment of its interconnections and parameters. tape storage of these
data, and automatic readout: both numerical and graphical. Although slower than
newer electronic equivalents, nonetheless for a 19-integrator problem which was run
on it in 1945, a thoroughly non-linear problem by the way, the analyzer time scale
was only twice as slow as the real scale for the remotely controlled glide vehicle
which was being simulated. The disc integrators of this machine were things of
beauty, with accuracies approaching, and resolution exceeding, 5 decimals. They
could integrate with respect to dependent variables, thus enabling multiplication
with only two integrators, logarithms without approximation, and so on. Integrators
of this same general type were also applied in astronomical and military computing
devices, in which less elaborate but still legitimate differential equations were em-
bodied and solved. This sort of equipment inspired many of the electronic analog
devices which followed, as well as the digital differential analyzers which have
come much later. Although the electronic integrators of analog equipment prefer
time as the direct variable of integration, they have shown extreme flexibility of
operating speed. One imagines the mechanical discs of the older analyzers running
at millions of rpm trying to keep up with their progeny!
The disc integrators of the differential analyzer worked without feedback, as did
its other basic parts. Where then did feedback appear in these analyzers? In the
differential equations acted out within it. Any equation requiring solution involves
at least one causal loop. But for feedback in its more exuberant forms we nominate
the next discipline to be considered, namely automatic controls.
Regulatory mechanisms such as those which are found in industrial control sys-
tems have been around for a long time. Roughly in historical sequence, they have
been mechanical, hydraulic, pneumatic, electric, and electronic. Translating as they
do from the unbalance or error in a controlled condition to the manipulation which
is intended to reduce that unbalance, they close a feedback loop which includes
some sort of plant. In typical cases these mechanisms have embodied mathematical
laws with continuous fidelity, and in order to attain fidelity they have resorted to
internal feedbacks precisely analogous to those employed in a modem amplifier. It
may not be widely known, particularly among the younger computing set, that this
sort of local feedback was applied in standard controller mechanisms of the twen-
ties and even earlier. These antecedent regulatory devices qualify as DC feedback
and even null-seeking at two distinct levels, and with mathematical capabilities, it
is not difficult to trace the logical paths of evolution from these devices to analog
computing as it is now enjoyed. Furthermore it is not uncommon in the thirties to
build simulators embodying convenient models of plants, into which the real regu-
latory mechanism could be connected. Both developmental and educational pur-
6