IEEE Solid-States Circuits Magazine - Spring 2020 - 25

Discussion on Figures of Merit

Power Consumption (mW)

In addition to key LNA performance
metrics, one or two figures of merit
(FOMs) are often reported in LNA
publications. More than 30 different LNA FOMs were identified while
compiling the database.
Typically, FOMs are reported at
the end of a paper in a summary table
detailing the performance metrics of

a particular LNA and other relevant
designs. An FOM can serve different
purposes; for example, it can show
how well a given LNA performs relative to other LNAs or how well its
design strategy performs relative to
other design strategies. Because of
the sheer number of different performance metrics, the former strategy
is typically adopted for comparing

0.1

10
Frequency (GHz)

16 nm
45 nm
180 nm
500 nm

22 nm
55 nm
200 nm
Trend

28 nm
65 nm
240 nm

100

32 nm
90 nm
250 nm

40 nm
130 nm
350 nm

FIGURE 3: CMOS LNA power consumption versus center frequency. The trend line indicates
an ~1.7 # increase in power consumption per decade of frequency.

10
Noise Figure (dB)

LNAs are designed to have similar
gains irrespective of technology or
frequency. These gains are largely
determined by the LNA load and, for
some LNAs, the amount of feedback
used for input power matching. The
gain can be increased by cascading
stages, which would normally have
only a marginal effect on power consumption, due to the relaxation of
the 50-Ω matching requirement, and
a relatively low effect on the noise
figure. While the gain-bandwidth
product of older technologies noticeably limits their operation at higher
frequencies, the gains of high-frequency LNAs in older technologies
remain similar to other designs. In---
terestingly, our analysis also shows
low correlations between gain and
power consumption, gain and noise
temperature, and gain and the output third-order intercept point (OIP3).
This final point suggests that, unless
extra stages are used to significantly
increase the gain, extra stages are
not normally expected to limit OIP3
for LNAs exhibiting typical gains.
We next plot CMOS LNA power
consumption, Pdc versus center frequency, f0, in Figure 3. A trend line is
plotted to identify an ~1.7# increase
in power consumption per decade
of f 0 . It is also visible that some
LNAs consume very low power. Conversely, some high-power-consuming LNAs achieving high linearity
also exist.
Finally, Figure 4 shows the noise
figure frequency trends for all LNAs
discussed in this survey. With the
exception of indium phosphide (InP)
and cryogenically cooled (cryo) LNAs,
the trend lines are nearly parallel but
slightly diverging for silicon germanium (SiGe), gallium nitride (GaN),
and gallium arsenide (GaAs) LNAs.
This is also true to some extent for
CMOS LNAs, which constitute by far
the largest number of designs in
this survey. Since uncooled InP LNAs
form the smallest LNA group in the
survey, it is possible that their relatively steep trend-line slope is due
to their comparatively small number
of entries.

8
6
4
2
0
0.1

10
Frequency (GHz)

CMOS
CMOS Trend
InP
InP Trend

SiGe
SiGe Trend
GaN
GaN Trend

100

GaAs
GaAs Trend
Cryo
Cryo Trend

FIGURE 4: LNA noise-figure trends with frequency for different semiconductor t-echnologies.
Cryo: cryogenically cooled.

IEEE SOLID-STATE CIRCUITS MAGAZINE

S P R I N G 2 0 2 0

IEEE Solid-States Circuits Magazine - Spring 2020

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Spring 2020