Professional Documents
Culture Documents
01618597
01618597
4, APRIL 2006
1687
I. INTRODUCTION
HE ALLOWANCE of the FCC regarding frequencies between 310 GHz for ultra-wideband (UWB) applications
has led to an increased level of interest and scope of research
on this band and its various applications. The availability of
such high bandwidth would allow higher data throughput up to
500 Mb/s over short distances, which is desirable for high-definition television (HDTV) and other wireless multimedia applications. Apart from high data rates, the other compelling features of UWB would be potentially lower cost and higher level
of integrations.
Several different approaches have been proposed to establish a universal standard for such UWB applications [9], [10].
Whether it is a pulse-based system or a multiband orthogonal
frequency-division multiplexing (OFDM) system, the basic requirement of the UWB transceiver is a wideband low-noise amplifier (LNA). The LNA needs to cover a wide range of frequencies from 3.1 to 10.6 GHz, along with exhibiting low noise
figure and low power consumption.
1688
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 54, NO. 4, APRIL 2006
(3)
where
(4)
input power match due to the fact that its input impedance is
mainly set by frequency invariant terms, given as
(1)
The small-signal transconductance
is a frequency-invariant term. Thus, the overall input impedance is set by the
term. As
is usually a resistor, the input
matching for maximum power transfer can be achieved over a
very wide frequency range. However, the feedback resistance
controls input impedance as well as a 3-dB bandwidth
and noise performance, thus requiring careful optimization of
its value to achieve low noise performance with a wideband
operating frequency range.
B. Noise Figure
The UWB specification requires the RF front-end to cover
frequencies from 3 to 10 GHz with a low noise figure and high
gain for a low dc power dissipation. Even though the resistive
feedback is generally known for its degradation in gain and
noise performance in tradeoff for the extended operating bandwidth, further analysis is required to facilitate for optimal design
(2)
The noise factor can be expressed in terms of
and the
source admittance [16]. The noise performance of an amplifier
is determined mainly by its minimum noise factor and noise
contribution that occurs when input source admittance is different from its optimum admittance. Due to its wideband inputmatching characteristics of a resistive feedback, noise contribution from the second term in (2) can be made small compared
, and low noise performance over a wide bandwidth can
to
.
be achieved by minimizing
The small-signal noise model of the resistive feedback SiGe
amplifier is shown in Fig. 2, where is the base resistance due
to current crowding effect,
is the baseemitter junction cais basecollector juncpacitance, diffusion capacitance is
and
are load and feedback impedtion capacitance, and,
ances, respectively.
(5)
As
increases, the frequency that defines a transition from
a white noise behavior to a 10-dB/decade increase of minimum
is 100 GHz
NF also increases. Suppose, for instance, that
and is 200: the transition occurs around 7 GHz without the
resistive feedback from (5). The transition frequency can be
increases. However, the reasonable transition
increased as
frequency is around 10 GHz for our purposes, considering
operating bandwidth and noise performance for a given bias
condition.
As derived in Appendix A, the expression of optimum source
susceptance for the low noise matching is
(6)
(7)
where
PARK et al.: ANALYSIS OF UWB SiGe HBT LNA FOR NOISE, LINEARITY, AND MINIMUM GROUP DELAY VARIATION
1689
(8)
Therefore, the required inductance
is
(9)
A resistive feedback reduces the optimum source susceptance
and, thus, increases the required input inductance for
noise matching compared to the case without the feedback or
the case with inductor degeneration, which is dominantly used
in conventional narrowband LNA design.
One advantage of noise matching at the high frequency is that
it can reduce the required input inductance. By matching noise
at the high-end frequency, this small increase in inductance can
be compensated and the required input inductance can be reduced. As a result, the overall noise performance can be improved because a high-quality inductor can be used. Therefore,
input noise matching at the high-end frequency will ensure the
low noise performance of the amplifier over a wide bandwidth.
(10)
where
and
are the gain and the phase components
of the transfer function, respectively.
Phase delay and group delay are expressed as follows:
Phase delay
(11)
Group delay
(12)
(13)
1690
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 54, NO. 4, APRIL 2006
where
Fig. 5. Effect of the input inductance on the group delay variation.
(14)
and
(15)
for the given circuit is expressed
The small-signal gain
in (13). The equation can be divided into two parts. One is the
from input
to point , and
voltage transfer function
from point to point . Intuthe other is the gain term
itively, since group delay is the derivative of the phase of the
transfer function, any resonance in the signal path will contribute distortion in the group delay. For CMOS design, delay
and
should be considered beperformance both in
cause of its relatively high parasitic capacitance in the drain.
However, the SiGe HBT has quite small parasitics in the colcan be negligible
lector, and the delay contribution from
as long as the inductance peaking with the parasitic capacitances
occurs out of the operating bandwidth. The main contribution
comes from the Miller effect. Because the input capacitance becomes large due to the Miller effect, it can resonate with the
input inductance. Therefore, the critical part for group delay
, and the input can be approximated
variation comes from
as in Fig. 4 as
(16)
where
(17)
(18)
(19)
(20)
(21)
PARK et al.: ANALYSIS OF UWB SiGe HBT LNA FOR NOISE, LINEARITY, AND MINIMUM GROUP DELAY VARIATION
tive load is quite helpful since it can boost the gain as the frequency increases while the intrinsic gain of the device decreases
as the frequency increases. The concern of the load inductance is
that it can distort the group delay variation, as discussed in [4].
As can be seen in (20), the load impedance also affects the
group delay variation by reducing . Therefore, the maximum
value of the load impedance which can minimize the group
delay variation exists.
In order to make sure of the flat group delay variation, the folis high-end frequency, which
lowing equation holds, where
is 10 GHz in UWB applications:
1691
(22)
is related to the load inductance
as in (17), we
Since
can obtain the following expression by solving (22):
(23)
In the design,
is set mainly for linearity in the SiGe device
is set, then the value of
considering its bias current. Once
load inductor can be determined using (23), which guarantees
low group delay variation over the entire bandwidth.
The calculated value from (23) can be used as an initial value
for the design. Since the delay variation not only depends on
but also
, the calculated value for load inductance can be
is less than 0.577.
increased as long as
V. LINEARITY ANALYSIS
Linearity behavior in the wideband is different from conversional narrow band circuit design. In narrowband applications,
the receiving bands are usually selected by a preselect filter,
eliminating high interferers from other applications. Therefore,
their linearity considers only their input 1-dB compression
point (P1dB) and input third-order intercept point (IIP3), which
cannot be removed by preselect filter.
The challenge in the UWB system is that the frequency band
allowed for system overlaps from other existing applications,
which are possibly near the range of the UWB system. The
major interference would come from 802.11a, for instance,
since it shares its frequency bands with the UWB bands. Even
though the transmitted and received signal strength is very low
compared to the conventional narrowband system, the linearity
of the LNA cannot be ignored since there is virtually no control
over the high interferers such as from 802.11a, which might
cause the RF front-end of the UWB system to saturate.
Generally, the linearity improves as current consumption increases. Because the power consumption of the LNA should be
minimized for UWB applications, there is a motivation to analyze the design concerning its linearity performance with low
power consumption.
Volterra-series analysis is performed on the LNA to gain an
insight into its linearity behavior in wide bandwidth. In this analysis, it is assumed that the input signal is very weak such that
(24)
is the th power of the voltage source signal and
where
is the Volterra-series coefficient, which is a linear function of
number of frequencies. The operator o indicates that each
is changed by the magnitude and
frequency component of
[15].
phase of
The analysis is performed on the resistive-feedback LNA
shown in Fig. 1 with the device size of 0.2 m 15 m, and the
1692
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 54, NO. 4, APRIL 2006
Fig. 8. (a) Circuit schematic of the implemented UWB LNA on SiGe BiCMOS process. (b) Chip photograph of the UWB LNA.
is
IMD
(25)
The details of procedure to derive a Volterra-series equation can
be found in [13][15].
For its wideband linearity behavior, the expression of IMD
can be divided into the three parts. is related to the overall
linearity performance over the operating bandwidth, which
mainly can be improved by consuming large amount of current.
shows that IMD can be reduced by canceling
with ,
and base inductance
.
which includes source impedance
At low frequency, the cancellation effect is negligible. However,
the amount of cancellation increases as the frequency increases
and, therefore, the IMD performance generally improves with
also suggests that the feedback resistance and
frequency.
load impedance should be as large as possible to minimize
IMD .
Since the load impedance is related to the group delay variation in this design, the value of the load inductor can be maximized as long as its degradation in group delay variation is
negligible.
determines its wideband linearity behavior. It can be defined as a multiplication factor because its value, which is less
(26)
Note that (26) shows that the frequency where the maximum
degradation occurs for its linearity performance is related to the
base inductance as well. Since the base inductance is related to
the noise and group delay performance of the LNA, the peak
degradation in linearity cannot be avoided for the UWB applications. However, by predicting its maximum degradation frequency in linearity, we might be able to move the frequency
point to the unused frequency bands as discussed in [9] and [10].
For the UWB LNA design, this peak degradation should be considered, and the linearity specification must be satisfied.
VI. CIRCUIT DESIGN
The base resistance is the main factor that increases
of the device, and it is related to the emitter width. The device
with smaller emitter width reduces
especially at high
PARK et al.: ANALYSIS OF UWB SiGe HBT LNA FOR NOISE, LINEARITY, AND MINIMUM GROUP DELAY VARIATION
1693
Fig. 11. Measured and simulated noise figure and group delay of the LNA.
Fig. 9. Measured and simulated S 21 and S 11 of the LNA.
Fig. 12. Measured IIP3 and ICP of the LNA over the frequency range.
Fig. 10. Measured and simulated S 12 and S 22 of the LNA.
1694
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 54, NO. 4, APRIL 2006
TABLE I
COMPARISON WITH PREVIOUS PUBLISHED WORKS
band, not only does the value of the inductor decrease but the
noise figure within the operating bands decreases because maximum noise figure would be at the high-end frequency.
Because the quality factor of the noise-matching inductor affects the overall noise performance, a transmission-line inductor
of 0.66 nH with a of 20 is used in the implementation.
is selected to achieve a best IIP3 perforLoad resistance
performs an important role in determining IIP3 of
mance.
can be calculated using bias
the device itself. The value of
current and supply voltage to bias the device at its high toleris determined, the value
ance point to the distortion. Once
of the load inductor that has a minimal effect on the group delay
variation is obtained.
PARK et al.: ANALYSIS OF UWB SiGe HBT LNA FOR NOISE, LINEARITY, AND MINIMUM GROUP DELAY VARIATION
Suppose that
as (3).
1695
, then
can be expressed
APPENDIX B
(27)
From (24),
,
pressed as follows [16]:
, and
can be ex-
(28)
and
are the Fourier transforms for noise
where
voltage and current defined by [18]
Using the -parameter equation described in [13], The ex,
can be derived as follows.
pression for
Supposing that
, we have
(36)
(37)
(38)
(29)
(30)
(31)
where
,
, and
From the above equations, we can derive following equations:
(39)
(40)
(41)
(32)
(42)
(34)
and (33), shown at the bottom of this page. Using (32) and (33),
is expressed as (35), shown at the bottom of this page.
(43)
(33)
(35)
1696
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 54, NO. 4, APRIL 2006
where
and
(44)
(45)
By substituting (42)(45) into (36)(38), we obtain the following equations:
(46)
(47)
(48)
(49)
ACKNOWLEDGMENT
The authors would like to thank A. Joseph, D. Harame,
D. Herman, J. Dunn, B. Meyerson, and the IBM SiGe team for
their contributions.
REFERENCES
[1] J. Lee and J. D. Cressler, A 310 GHz SiGe resistive feedback
low noise amplifier for UWB applications, in IEEE RFIC Dig., pp.
545548.
PARK et al.: ANALYSIS OF UWB SiGe HBT LNA FOR NOISE, LINEARITY, AND MINIMUM GROUP DELAY VARIATION
[30] Y. Park, C.-H. Lee, J. D. Cressler, J. Laskar, and A. Joseph, A very low
power SiGe LNA for UWB application, in IEEE MTT-S Int. Microw.
Symp. Dig., Jun. 2005, pp. 10411044.
John D. Cressler (S86A91SM91F01) received the B.S. degree in physics from the Georgia
Institute of Technology (Georgia Tech), Atlanta, in
1984, and the M.S. and Ph.D. degrees in applied
physics from Columbia University, New York, in
1987 and 1990, respectively.
From 1984 to 1992, he was a Member of the
Research Staff with the IBM Thomas J. Watson
Research Center, Yorktown Heights, NY, and from
1992 to 2002 on the faculty at Auburn University, Auburn, AL. In 2002, he joined the faculty
at Georgia Tech, where he is currently the Byers Professor of Electrical
and Computer Engineering. His research interests include: SiGe/strained
Si devices and technology, mixed-signal circuits built from such devices,
radiation effects, cryogenic electronics, device-to-circuit interactions, reliability physics, device-level simulation, and compact circuit modeling. He has
authored or coauthored over 300 papers related to his research. He coauthored
Silicon-Germanium Heterojunction Bipolar Transistors (Artech House, 2003),
authored Reinventing Teenagers: The Gentle Art of Instilling Character in Our
Young People (Xlibris, 2004), and edited Silicon Heterostructure Handbook:
Materials, Fabrication, Devices, Circuits, and Applications of SiGe and Si
Strained-Layer Epitaxy (CRC, 2005).
Dr. Cressler was an Associate Editor for the IEEE JOURNAL OF SOLID-STATE
CIRCUITS (19982001), the IEEE TRANSACTIONS ON NUCLEAR SCIENCE
(20022005), and currently for the IEEE TRANSACTIONS ON ELECTRON
DEVICES. He has served on the Technical Program Committees of the IEEE
International Solid-State Circuits Conference (19921998, 19992001),
the IEEE Bipolar/BiCMOS Circuits and Technology Meeting (19951999,
1697