Figure 1. Proposed Balun-LNA with capacitive cross-coupling technique
A 2 GHz - 8.7 GHz Wideband Balun-LNA with Noise Cancellation and Gain Boosting
T. Chen
1, S. Rodriguez
1, E. Alarcon
2, and A. Rusu
11
School of Information and Communication Technology KTH Royal Institute of Technology
Stockholm, Sweden tingsu@kth.se
2
Department of Electronics Engineering UPC Universitat Politecnica de Catalunya
Barcelona, Spain
Abstract— A wideband Balun-LNA covering the operation frequency range of magnetic tunnel junction Spin Torque Oscillator is presented. The LNA is a combination of common- source and cross-coupled common-gate stages, which provides wideband matching and noise cancellation, as well as gain boosting. The internal feedback introduced by the cross-coupling allows an additional degree of freedom to select transistor sizes and bias by decoupling the impedance matching, noise, and gain imbalance trade-offs which are present in similar topologies. Two LNAs using the proposed technique are designed in 65nm CMOS.
The LNAs have a simulated bandwidth of 2 GHz - 8.7 GHz, gain of 16 dB, IIP3 of -3.5 dBm, and NF < 3.8 dB while consuming 3.72 mW from a 1.2 V power supply.
I. I NTRODUCTION
The multi-standard radio receivers [1-6] as well as emerging spin torque oscillator (STO) technology [7] demand high performance low noise amplifiers (LNA) operating across a wide frequency range. The common-gate (CG) stage has been frequently used in wideband LNAs for its broadband input impedance matching. However, it suffers from poor noise performance and low voltage gain. To overcome these shortcomings, common-source (CS) stage is employed in combination with CG to generate an anti-phase signal, forming an CG-CS topology [1] [2]. The CG-CS topology provides noise and distortion cancellation, broadband input impedance matching and simultaneous gain balancing. The conventional CG-CS topology [3] has trades-offs between power consumption, noise figure (NF), bandwidth, voltage gain and gain imbalance. An improved CG-CS LNA using local feedback to boost the CG transconductance has been recently proposed in [4]. Although this local feedback brings gain boosting and good linearity, the parasitics of the additional transistors degrade the bandwidth. Therefore, it is hard to fulfill the bandwidth requirement for wideband applications. To alleviate these problems, this paper proposes a new CG-CS Balun-LNA circuit using capacitive cross-coupling technique, targeting the STO applications.
STO is a novel tunable nanoscale microwave integrated current-controlled oscillator (CCO), which benefits from its wide tunability, miniature size, high operation frequency, high integration level, and high quality factor (Q) [7]. Nevertheless, its drawback is that it provides very limited output power and therefore requires low-noise amplification in order to become
usable. The proposed Balun-LNA is required to preamplify the output signal of a magnetic tunnel junction (MTJ) STO so that a single local oscillator (LO) can cover a large part of the ultra- wideband (UWB) spectrum from 3 GHz - 8 GHz. The MTJ STO used as basis for this paper has an output impedance close to 50 Ω [8] and therefore requires 50 Ω transmission lines and termination in order to maximize the power transfer and avoid undesirable reflections. In addition to the low noise performance and wideband impedance matching (small S
11) in the band of interest, the requirements for this LNA are wideband operation (3 GHz - 8 GHz), high gain (at least 10 dB) and low power consumption. The proposed LNA is not directly connected to an output buffer stage since it will be later integrated with the limiting amplifier chain for further amplification.
This paper is organized as follows. Section II presents the topology of the proposed Balun-LNA. Section III provides the theoretical analysis and design of the proposed LNA. Section IV discusses the simulation results while Section V draws conclusions.
II. P ROPOSED BALUN-LNA
The topology of the proposed Balun-LNA is shown in Fig. 1. The Balun-LNA consists of a CS stage (M
1) combined with a cross-coupled CG stage (M
2, M
3) [5]. The impedance at the source of M
2is designed to match the source. Biasing of CG stage can be done by using either an NMOS current source or an inductor. Transistors M
0-3are biased by current mirrors,
This work was supported in part by the Swedish Research Council (VR) and the Spanish Ministry of Education (2008-2011 I+D+I National Program)
© 2012 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses,
in any current or future media, including reprinting/republishing this material for advertising or promotional
purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted
component of this work in other works.
Figure 2. Small signal model of cascoded CS stage which are not shown in the figure. The voltage drop at R
1has
the same sign as the source and produces the output voltage V
out+. The voltage drop at R
2has opposite sign due to inversion caused by M
1and produces the output voltage V
out-.When the amplitude of the outputs is equal, the amplifier behaves as a single-ended to differential balun converter, allowing differential processing of the signal on the following stages.
Transistor M
1is cascoded by M
3, hence improving its voltage gain. This arrangement also reduces the Miller effect on M
1and therefore helps to enhance the bandwidth. The cross-coupled transistors introduce a dual feedback loop into the circuit which allows more design variables to enter into the expressions of input matching, gain, and noise performance. The next section describes how these feedback loops enable more degrees of freedom that can be exploited in order to optimize the LNA performance.
III. T HEORETICAL D ESIGN - ORIENTED A NALYSIS A. Gain Boosting
The voltage gain from V
Sto V
out-is found by using the small signal equivalent circuit of Fig. 2. Since the drain-source resistance can be as low as a few hundred Ohms, the effect of channel length modulation of M
1needs to be accounted for this analysis.
The voltage gain V
out-/ V
Scan be expressed as:
out- m1 1
m3 2
S m3 1
1 1
o o
V g r
A g R
V g r
⋅ +
= − ⋅ ⋅ ⋅
⋅ +
(1)
where g
m1, g
m3, and r
o1are the transconductance of M
1, M
3, and the drain-source resistance of M
1respectively. Α is a factor that accounts for the voltage divider at the input, and that takes the value of 0.5 when the source impedance is matched.
The voltage gain calculated from V
Sto V
out+can be also found using a similar procedure and its expression is:
out+ m1 1
m 2 1
S m3 1
1 1
o o
V g r
A g R
V g r
⋅ +
= ⋅ ⋅ ⋅
⋅ +
(2)
where g
m2is the transconductance of M
2. Compared to the CG stage where the gain is roughly A g
m2R
1, (2) shows that the gain is boosted by a factor (g
m1·r
o1+1)/(g
m3·r
o1+1) due to the feedback loop.
Accordingly, the total voltage gain of the proposed LNA is:
out+ out- m1 1
m2 1 m3 2
S m3 1
1 ( )
1
o v
o
V V g r
A A g R g R
V g r
− ⋅ +
= = ⋅ ⋅ +
⋅ +
(3)
The condition for balancing the differential outputs can be found by equating (1) and (2), which results in the following relationship:
m3 1
2 m 2
g R
R = g (4)
When the differential gain is balanced, the total voltage gain can be simplified to:
m1 1 m1 1
m 2 1 m3 2
m3 1 m3 1
1 1
2 2
1 1
o o
v
o o
g r g r
A A g R A g R
g r g r
⋅ + ⋅ +
= ⋅ ⋅ ⋅ = ⋅ ⋅ ⋅
⋅ + ⋅ +
(5)
where it can be appreciated that the boosting factor increases the total gain without requiring additional power consumption.
This factor is one of the main advantages of this cross-coupled configuration.
B. Input Impedance
The input impedance R
inof the proposed LNA is given by:
m 3 o1 in
m1 o1 m 2
1
( 1)
g r
R g r g
⋅ +
= ⋅ + ⋅
(6)
Proper matching requires R
in= R
S. Under this condition, g
m2can be expressed by:
m 2
m1 o1
S
m3 o1
1 1 1
g g r
g r R
= ⋅ +
⋅ + ⋅
(7)
In this case, g
m2appears to be inversely proportional to the boosting factor. Therefore, a proper combination of g
m1, g
m3, and r
o1allows small M
2transistor sizes and biasing currents, thereby resulting in small parasitics and low power consumption.
C. Noise/Distortion Cancellation
In a CG stage which is matched to a given source impedance, half of the drain noise current flows directly to R
1resulting in a minimum noise factor of (1+ γ), where γ represents the drain noise coefficient. The proposed topology, however, belongs to a family of CG-CS LNA that provides a cancellation mechanism for suppressing this noise. The differential output noise voltage generated by i
n,M2, the drain noise current of M
2, is:
2 2
nout+,M nout-,M
m1 1
1 S m3 2 n, 2
m3 1
[ 1 ]
1
o
M o
V V
g r
A R R g R i
g r
− =
⋅ +
− ⋅ − ⋅ ⋅ ⋅
⋅ +
(8)
which can be cancelled provided that the following condition is satisfied:
m1 1
1
m3 S
2 m3 1
1 1
o o
g r
R g R
R g r
⋅ +
= ⋅ ⋅
⋅ +
(9)
This condition is independent of A. By replacing (9) in (7) it is
possible to obtain the same expression as (4), which means
that noise cancellation, matching, and gain balancing are
0 10 20 30 40 50 60 2
2.5 3 3.5
gm1 (mS)
NF (dB)
R1a=500 Ohms R1b=350 Ohms R1c=200 Ohms
Figure 3. Relationship between g
m1, R
1and noise
2 3 4 5 6 7 8 9
−30
−20
−10 0 10 20
Frequency (GHz)
S11, Av (dB)
Gain1 (TT, inductor) Gain2 (TT, NMOS) Gain3 (FF, NMOS) Gain4 (SS, NMOS) S11 (TT, inductor) S11 (TT, NMOS)
Figure 4. Simulated voltage gain A
vand input return loss S
112 4 6 8
2 3 4 5 6
Frequency (GHz)
NF (dB)
NF1 (TT, inductor) NF2 (TT, NMOS) NF3 (FF, NMOS) NF4 (SS, NMOS) NF5 (Calculation)
Figure 5. Simulated Noise Figure concurrently achieved. Moreover, the non-linear distortion
generated by M
2can be cancelled under the same conditions.
The other noise contributors are M
1, the load resistors R
1and R
2, and the biasing of M
2. For simplicity, the following analysis disregards the noise of this biasing; however, its effect will be studied in the later examples in the next section. The relationship between g
m1, g
m2, g
m3, R
1, R
2and the noise performance can be expressed by the noise factor F given by:
1 2 1 2
S
2 2 2 2
nout , nout , nout , nout ,
2 2
n ,
1
M M R RR v
F
A
ν ν ν ν
ν
+ + +
= +
⋅
(10)
where ν
nout,M1, ν
nout,M2, ν
nout,R1and ν
nout,R2, represent the differential output noise voltage of M
1, M
2, R
1and R
2. The noise factor as a function of the small signal parameters and noise sources can be expressed as:
1
S
2 1 2
S
2 1 2 2
m 2 1 m3 2 n ,
m3 1
2 2
n ,
2 2 2 2
m1 1
1 S m3 2 n , nout , nout ,
m3 1
2 2
n ,
1 ( ) ( )
4 1
1
1
1 [ ]
4 1
o
M o
R v
o
M R R
o
R v
g R g R r i
g r F
A g r
R R g R i
g r
A ν
ν ν
ν
+ ⋅ ⋅
⋅ +
= +
⋅
⋅ +
− ⋅ ⋅ + +
⋅ +
+
⋅
(11)
where i
n,M1, i
n,M2are drain noise currents of M
1, M
2. When (9) is satisfied, the noise of M
2is cancelled and the total noise factor simplifies to:
S S 2
1 m1
2 1 12
S m1 o1
1 1
( )
R R R F g
R R
R g r γ α
⋅
= + ⋅ + +
⋅ +
(12)
where γ
1is the drain noise coefficient of M
1and α represents the transconductance degradation in the short-channel regime.
R
1and R
2are chosen much larger than R
S, causing that M
1governs the noise contribution. From (12), it can also be seen that as g
m1increases, a lower F is achieved. Furthermore, according to (5), increasing g
m1boosts the differential gain simutaneously. Fig. 3 illustrates the relationship between R
1, g
m1and NF (NF=10
log(F)), by using (12) and assuming γ
1= 1, α = 0.8; r
o1= 325 Ω is from simulation. Considering (4) and (5), the balanced differential gain is proportional to R
1. As it can be seen, the gain and noise improvement in this case can be concurrently achieved, which is opposite to [6] where larger load resistors degrade the noise performance.
IV. S IMULATION R ESULTS
In order to validate the previous theoretical analysis, two LNAs were designed using 65 nm CMOS technology. The first LNA uses an NMOS current source M
0to bias M
2, whereas the second LNA uses an on-chip inductor. Based on previous analysis M
1, M
2, and M
3take the sizes 50.4um/65nm, 20um/65nm, and 28um/65nm, respectively. M
1, M
3are biased at 1.88 mA which results in g
m1, and g
m3of 20 mS and 15 mS, repectively. M
2is biased at 1.22 mA so that g
m2is approximately 10 mS. In the case when M
2is biased with a current source, M
0has a size of 40um/500nm. R
1and R
2are
set to 500 Ω and 325 Ω, respectively. Load capacitances C
Lequal to 20 fF representing a wideband limiting amplifier stage have been included at the output. Finally, a 1.5 nH bonding wire and 120 fF capacitances to account for the pads and ESD protection were added at the input. Using the theoretical derivation presented in section III and the design values of the proposed LNA, the voltage gain can be calculated as 16.1 dB when the input impedance is perfectly matched to R
S. The NF can be estimated as 2.26 dB.
Fig. 4 shows the simulated voltage gain A
vand return
losses S
11for both LNAs. In both cases, the gain matches very
well the theoretical result using (3). For the fast - fast (FF)
corner simulation at -25
oC and (VDD+10%), the proposed
LNA (current source biasing case) exhibits 17.58 dB gain with
2 GHz - 8.1 GHz bandwidth. For the slow - slow (SS) corner
at 85
oC and (VDD-10%), the gain is 14.3 dB and the
bandwidth is 2 GHz - 9 GHz. The requirements of gain and
bandwidth can still be fulfilled in the worst cases. Moreover,
−60 −40 −20 0 20
−200
−150
−100
−50 0 50
RF input (dBm)
Output Voltage (dB)
IM3 (Inductor) IM3 (NMOS)
Vout
IIP3(NMOS)
−3.5dBm
IIP3(Inductor)
−1dBm P1dB
Figure 6. Two-tone IIP
3simulation results (fo = 5 GHz) (TT corner)
Figure 7. Monte Carlo simulation results for process variation (TT corner)
TABLE I. C
OMPARISON WITH PREVIOUSBALUN-LNA
DESIGNSTech.
(nm) VDD (V)
BW (GHz)
Gain (dB)
NF (dB)
IIP
3(dBm) S
11(dB) Power (mW) This
design *
65 1.2 2-8.7 16
1)<3.6 -3.5 -9 3.72 65 1.2 2-8.7 15.5
2)<3.8 -1 -14 3.72 [2] 130 1.2 2.1 5.2** 3 +10.5 -14 12.6 [3] 65 1.2 .2-5.2 15.6 <3.5 0 - 21
[4] 130 1 .2-3.8 19 3.4 -4.2 -9 5.7
[4] 130 0.85 .2-3.8 14.8 4.1 -3.8 -9 3.2
* For this design, two biasing methods 1). NMOS and 2). Inductor are used, corresponding to Fig.1.
** measured S21