IEEE Solid-States Circuits Magazine - Summer 2023 - 66

RCA developed the RX shown in Figure
9(a), where dual-gate PMOS transistors
Q1-Q7 were grown on the same
substrate [14]. Transistor Q1 acts as a
cascode low-noise amplifier (LNA),
and Q2 as a mixer. Figure 9(b) shows
the chip photograph. Targeting military
FM radios in the range of 30-
76 MHz, the chip takes advantage of
diode-connected transistors Q3-Q7
to define the bias points of Q1 and
Q2. (Note that VS is a positive supply
voltage.) In fact, a closer examination
reveals that Q6 and Q2 form a
current mirror. RCA's report also recommends
that the LO be integrated
on the same chip in future work.
Unfortunately, RF CMOS would lie
dormant for about two decades.
Mixing: Nonlinearity Versus
Time Variance
Much of the early work on RF systems
considered mixing as the time-domain
multiplication of two waveforms
in a nonlinear circuit. Approximated
by a polynomial of the form
y ( t ) =
aa a++ +g,
12 ()
xt xt xt
()
2
3
3
()
the characteristic would sense the
sum of the RF signal and the LO
and generate their product through
even-order nonlinearity.
In 1963, however, Read proposed
that " any device in which mixing
occurs can be represented by a linear
network with time-varying components "
[17]. In fact, Read assumed
that a transistor's current resembled
a square wave and hence computed
the conversion gain.
The significance of " time-variant "
mixers was not appreciated
until later in that decade. Over the
years, it was discovered that 1) the
RF port of a mixer must be linear to
withstand blockers or at least preserve
the information in an AM signal
and 2) the LO port must switch
abruptly and completely so as to
provide a high gain and low NF. One
could argue that the dual-gate mixer
in Figure 9(a) exemplifies such an
implementation, but [14] does not
teach these principles. These points
emerged only as the spectrum be -
came more crowded and RF interferers
more abundant. The need for a
mixer with one linear port and one
switching port was fulfilled by Bilotti
and Pepper [18], as explained in the
next section.
FIGURE 7: A CMOS mixer from 1968 [11].
(a)
(b)
FIGURE 8: (a) A CMOS oscillator from 1968 [12] and (b) its counterpart from 1994 [13].
66 SUMMER 2023
IEEE SOLID-STATE CIRCUITS MAGAZINE
Bilotti Versus Gilbert
In an article published in 1967,
Bilotti and Pepper present the circuit
depicted in Figure 10(a), recognizing
the need for complete
switching in the LO path [18]. They
assume a small signal level arriving
at the top transistors and apply
the LO to the bottom pair. In the
1968 International Solid-State Circuits
Conference (ISSCC), Bilotti
reports the same topology but with
the top devices acting as switches
and the bottom ones degenerated
for high linearity [Figure 10(b)] [19].
At the same conference, Gilbert
offers the topology shown in Figure
10(c), where both ports are linear
[20]. We note that Bilotti upholds
the two principles described in the
previous section and should be
credited for the double-balanced
active mixer used today. Bilotti also
explains that the double-balanced
arrangement cancels the LO component
at the output.
IEEE Solid-States Circuits Magazine - Summer 2023

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