Professional Documents
Culture Documents
Abstract-----In this paper, a novel architecture for wideband matching stage for UWB LNAs, Common Source (CS)
input impedance matching consisting of two common gate (CG) topology is usually avoided as it generally exhibits a
transistors is presented. One CG transistor is placed on top of the narrowband response. Obtaining wideband response requires
other in a current reuse fashion such that both transistors appear
additional reactive components placed before the CS transistor
in parallel at the input. As a consequence, the transconductance
requirement for input matching from each NMOS transistor is which actually serve as bandpass filter [4],[6]. Additionally,
reduced to half compared to a simple CG LNA without effecting noise characteristics of a CS based LNA vary directly with
the total gain. The proposed input matching technique achieves a frequency, thereby increasing Noise Figure (NF) at higher
large bandwidth and high gain with comparatively small power frequencies. Moreover, due to Miller's effect, addition of a
consumption. The designed UWB LNA is simulated using IBM cascode stage is required which requires more voltage
130nm CMOS process with Spectre RF. Post Layout simulation
headroom thereby increasing power. The CG LNA, on the
results depict S11 and S22 better than -6.5 dB and -15 dB
respectively. The gain and 3 dB bandwidth are 15 dB and 2.1 contrary, excellent reverse isolation properties inherently. It
GHz, respectively. The LNA demonstrates minimum NF of 3.7 has a slightly higher but relatively constant noise figure
dB at 3.53 GHz in the passband, input referred 1dBCP of −15.32 independent of both the operating frequency and bandwidth
dBm with IIP3 of -10.5 dBm. The proposed LNA consumes only [1-3]. Its input capacitance is easily resonated out by
2.6 mW with Vdd of 1.4 V. adding an inductor at its source while its transconductance
helps to fix the 50Ω input impedance match. These features
Keywords----- CMOS LNA; common gate (CG); current reuse; makes CG LNA superior in performance compared to CS
input impedance; inverter; UWB LNA; Wideband LNA. LNA for wideband operation. A wideband CG LNA work
without a band pass filter at the input thereby saving area and
makes LNA design simpler.
I. INTRODUCTION In a conventional CG LNA, width of input transistor and
The Ultra Wideband (UWB) technology, existing from 3.1- power consumption are governed by transconductance
10.6 GHz has emerged as one of the most popular, most required for input matching. In such LNAs, g of 20 mS is
researched and most promising wideband technology after theoretically required from input transistor for input matching
being declared commercial by the Federal Communications whereas values of 22-23mS are usually set to cater for
Commission (FCC) in 2002. The extremely high data rate reflection of load impedance at the input [1]. The high power
together with large bandwidth, high protection, low cost and consumption associated with this transconductance value can
low power makes it a suitable choice for various wireless be restricted using current reuse techniques which have proved
communication applications including IrDA, Wireless USB, to be very efficient for low power designs [7-8]. To design a
Bluetooth, Z-Wave, ZigBee, Body Area Network etc. low power CG LNA, we here propose a new input matching
Besides, it has also been widely used in Biomedical architecture with the goal to use lesser current and achieve a
applications and vehicular tracking systems. comparable gain as in standard CMOS CG LNAs. In our
UWB Low Noise Amplifier (LNA) design is a great proposed design, one CG transistor is placed at the top of
challenge because the stringent requirements of good input other such that the sum of transconductances of both
and output matching, reverse isolation, flat and high gain, low transistors provide the desired 50Ω matching. Thus input
noise figure, and high linearity shall all be met over a very matching is possible at approximately half the
large frequency spectrum. Literature is full of UWB LNAs transconductance/ current consumption value compared to
implemented using the ever shrinking CMOS technology [1-4] conventional CG LNA. The proposed LNA achieves
which has achieved fT as high as 280 GHz (45nm CMOS) [5]. comparable performance with low power compared to
While CG topology is successfully used as an input previously reported LNAs.
978-1-4673-9680-6/16/$31.00
Authorized licensed use ©2016
limited to: J.R.D. Tata Memorial Library Indian IEEE
Institute of Science Bengaluru. Downloaded on March 20,2024 at 20:33:41 UTC from IEEE Xplore. Restrictions apply.
2016 5th International Conference on Modern Circuits and Systems Technologies (MOCAST)
VDD
C6 L4 Vb4
C7 M6
Vb2 R4 Vb6
M2 (1 / gm 6 ) ro 5
R2 M4 R6
M5 ≅ (1/gm6)
C2 L3 M6
C5
C8 L5 RFOUT
C3 L 2 C4
RFIN M3 C9 Fig. 2. Simplified schematic showing the 50 Ω output impedance
Vb1 M1 R3 M5
R1
R5 Capacitor C5 also presents an ac ground over the entire
Vb3
C1 Vb5 bandwidth of interest. The RF signal after amplification
L1
through M1 and M2 is coupled through C4 and C7 to the gain
Input Matching Gain Stage Buffer
stage (M3 and M4) again in a current reuse fashion. This stage
Stage provides additional gain to the LNA. Besides current reuse, the
Fig. 1. Proposed current reuse common gate LNA with novel input placement of M3 and M4 in an inverter configuration helps to
impedance matching architecture improve the linearity which depends on latter stage of the
LNA. The output of this stage is coupled through C8 to the
The paper is divided as follows. In section II, main design
buffer stage which also isolates the dc bias between these
concept, mathematical equations and circuit parameters are
stages. Series resonance of this capacitor with L5 and another
presented. Layout design and post layout simulation results are
resonance of L5 with input capacitance of M6 assist in
presented along with comparison with recent LNAs in section
extending bandwidth towards higher frequency spectrum of
III and the paper is finally concluded in Section IV.
desired band. The gain (A) of the LNA can be expressed as in
II. CIRCUIT DESIGN (4).
A. Design theory = −( ‖ )( +
2 4) (4)
Fig. 1 shows schematic of the proposed LNA. In input
matching stage, C1 and C2 couple input signal to the sources of where r and r are the drain to source resistances of M1 and
CG transistors M1 and M2 respectively. M2 is stacked at top of M2 respectively. g , g are the transconductances of M3 and
M1 which allows them to share the same drain current. They M4 respectively. sL and sL are the impedances of L and L
both act as input impedance matching device and their input respectively. Noise Figure for the proposed LNA can be
impedance appear in parallel at the input node as in (1). The written as in (5).
input impedance of either M1 (Z ) or M2 (Z ) consists of
inverse of sum of respective transconductance (g for M1 and =1+ + (5)
g for M2) and the gate to source capacitance (sCgs for M1
and sCgs for M2) present at the source node (2) - (3). L1 and
L3 resonate out sCgs and sCgs respectively as in simple CG where noise of gain stage has been neglected in accordance
LNAs. Thus, real part of input impedance (Z ) looking into the with Frii's formula. In (5), γ is a noise parameter (coefficient of
LNA is equal to inverse of the sum of the transconductances of channel thermal noise) for deep submicron MOSFETS and its
M1 and M2 only i.e. = 1⁄( + ). value can be much greater than 1 for short channel devices.
This resistive component is used to match the LNA input ≌ / 0 and its value can be much less than 1 for short
impedance to the source resistance. channel devices [9].
M5 and M6 implement output buffer needed for
= ‖ (1) measurements with standard 50Ω lab equipment. This buffer
is not needed in integrated implementations where mixer is
placed on chip eliminating the requirement of 50Ω output
= (2)
matching. Here, inverse transconductance of M6 (1/g ) and
the output impedance of M5 (r ), appear in parallel at the
output node of the LNA. By adjusting g = 50 Ω and a high
= (3)
r , 1 g ‖r reduce to 1/g only, shown in Fig. 2. In this
way, output impedance matching is achieved through the
The gates of M1 and M2 are kept at ac ground with the help of buffer.
C3 and C6. L and L serve as load for M1 and M2 respectively The gate bias of all transistors is provided through current
and carry same bias current. The combination of inductors L , mirrors. Input matching stage is biased at 0.53 mA while gain
L and capacitor C5 make a third order LC-T network and stage carries 1.34 mA. Buffer is biased separately with a
provides adequate isolation between both M1 and M2. power supply of 1 V.
Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on March 20,2024 at 20:33:41 UTC from IEEE Xplore. Restrictions apply.
2016 5th International Conference on Modern Circuits and Systems Technologies (MOCAST)
-5
-10
S11 (dB)
-15
-20
Post Layout Simulation
Pre-Layout Simulation
-25
Fig. 3. Layout of the proposed LNA with an area of 1.1 x 0.67 mm2 4 2 5 63 7
Frequency (GHz)
excluding bondpads Fig. 4. Pre and Post Layout Simulated Input Reflection
Coefficient (S11) of the LNA
0
B. Circuit Parameters
From (1), g = g = 10mS, ideally, and simulations are -10
S22 (dB)
started with this value. After rigorous simulations and keeping
in view low power requirement, transconductance values are -20
adjusted to g = 9.3mS and g = 8.3mS respectively by
-30
setting M1 and M2 for gate widths of 40µm and 30µm Post Layout Simulation
respectively. With these values, S11 values as low as -15dB are Pre-Layout Simulation
-40
obtained in passband. Increasing gm's above the 2 3 4 5 6 7
Frequency (GHz)
aforementioned values only increase power consumption Fig. 5. Pre and post layout simulated output reflection coefficient
without much difference in S11. Hence above values are (S22) of the LNA
retained. 20
Inductors L , L are selected to be 7.6 nH each. They
resonate with gate to source capacitances plus other parasitic 10
capacitances present at source nodes of M1 and M2
S21 (dB)
Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on March 20,2024 at 20:33:41 UTC from IEEE Xplore. Restrictions apply.
2016 5th International Conference on Modern Circuits and Systems Technologies (MOCAST)
Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on March 20,2024 at 20:33:41 UTC from IEEE Xplore. Restrictions apply.