IEEE Solid-States Circuits Magazine - Summer 2023 - 31

structures. TCAD did not exist in
Shockley's day; that he developed a
compact model to understand and
explain his proposed device is simply
amazing.
Another proposal from that paper
still permeates our industry. For the
terminal names for his (theoretical,
not experimentally verified) " unipolar
field-effect transistor " :
The choice selected is " source "
for the electrode through which
the carriers flow into the channel,
" drain " for the electrode into
which the carriers flow out of
the channel, and " gate " for the
control electrodes that modulate
the channel.
Once the transistor was invented,
the advantages of being solid-state
(robustness, no need for a cathode
heater) were sufficient to pique interest
despite the high cost. However, to
move from being a research laboratory
breakthrough to being a commercial
success, transistors needed
to be used in products.
This meant circuit designers
needed to mentally understand how
transistors worked, and have some
form of capability to do " back of
the envelope " (SPICE would not arrive
for over two decades) calculations
to understand how their circuits
behaved, and how to properly design
and optimize component values and
biasing in their circuits.
The reference for this is the classic
Electrons and Holes in Semiconductors
(with the important subtitle With
Applications to Transistor Electronics)
by Shockley [2]. Many of us older
semiconductor engineers considered
this the " bible, " and cut our teeth
digesting it. And many early transistor
circuits were designed based
on the mental images (i.e., models)
espoused in that book.
However, the emphasis in [2]
is
more on the detailed physics of transistor
and pn-junction operation
than on the key aspects of device
behavior, in essence large-signal
behavior (from which dc operation
and small-signal behavior naturally
follow), that are critical to
IBE
B
FIGURE 2: Modern representation of the dc Ebers-Moll BJT model. bF = IC /IB in forward
operation and bR = IE /IB in reverse operation. Each of IBE and IBC have a form like (1).
IEEE SOLID-STATE CIRCUITS MAGAZINE
SUMMER 2023
31
IBC
The most incredible feature of this work was
that it developed a compact model to propose
a new form of transistor that did not exist.
understand transistor behavior to
enable circuit design.
A Side-Track Into Circuit Simulation
As mentioned previously, the concept
of a compact model was introduced
more than 10 years before
computer programs to analyze circuit
performance appeared. The original
purpose of the compact model
was to encapsulate the operational
features of a new device and to assist
engineers to design circuits that
employed that new device. A requirement
of these compact models was
that they had to be fairly simple and
any accompanying equations had
to be amenable to evaluation with
a slide rule or a calculator. Because
compact models were intended to
assist engineers with circuit design,
it is not surprising that these models
were schematic representations
of the device, as opposed to lengthy
equations. The classic example of
this type of compact model is the
Ebers-Moll BJT model, published
by Bell Laboratories researchers
J. J. Ebers and J. L. Moll in 1954 [6],
which is shown in Figure 2.
Compact modeling changed drastically
with the arrival of circuit analysis
programs in the 1960s, followed
by the evolution of circuit simulators
in the 1970s. Early circuit analysis
programs include the electronic
circuit analysis program (ECAP) from
IBM [11] and NET-1 from the Los Alamos
National Laboratory (CA, USA)
[12]. At about the same time, Jensen
published a paper describing his
implementation of the Ebers-Moll
model in ECAP for dc and transient
analysis [13].
Combining circuit simulators and
compact models had a significant
effect on the design community. On
the one hand, simplicity was not as
relevant, since it was no longer the
designer who had to evaluate the
device equations. Compact models
could therefore include greater complexity
to provide a more physical
representation of the device. On the
other hand, the designer became saddled
with the task of programming
compact models.
The early circuit analysis programs,
such as ECAP [11], required
the user to supply a FORTRAN subroutine
for each device model, and
the models were compiled and
loaded along with the circuit analysis
program. This was a daunting task
for circuit designers and device physicists
who were unfamiliar with computer
programming. Moreover, when
the University of California, Berkeley,
USA, included circuit simulation
as a part of circuit design courses
[14], it was clear that the simulators
had to be easy to use to work in the
EC
βFIBE - βRIBC
βRIBC - βFIBE
IEEE Solid-States Circuits Magazine - Summer 2023

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Summer 2023
