A Comparative Study of CMOS LNAs

Sherif A. Saleh, Maurits Ortmanns, and Yiannos Manoli

Chair of Microelectronics, Department of Microsystems Engineering - IMTEK,
University of Freiburg, Georges-Koehler-Allee 102
79110 Freiburg, Germany
Email: {Sherif.mohamed, Ortmanns, Manoli}@imtek.de
AbstractThree CMOS RF low noise amplifier circuits have
been designed and simulated. These LNAs are intended for use
in 402-405MHz Medical Implant Communication Service
transceivers. The inductively degenerated common source LNA
(CS-LNA) topology is currently popular because it achieves high
gain, low noise figure, and high linearity. In this paper cascode
LNA with inductive source degeneration, LC folded cascode
LNA topology and current reuse technique are used in a CMOS
CS-LNA with inductive source degeneration. The performance
target is to achieve a moderate gain and moderate noise figure
without overly degrading the linearity.

II.

DESIGN OF CASCODE LNA WITH INDUCTIVE SOURCE

DEGENERATION

A cascode LNA with inductive source degeneration

structure is widely used in receiver design. It is easier to
achieve input matching for both the power gain and the noise
figure. The channel width of the input transistor determines
the noise performance of the LNA because the dominant noise
source in CMOS devices is the channel thermal noise [3].
Since the biasing current is fixed, an optimum channel width
can be obtained for minimum noise contribution. Fig.1 shows
a single ended cascode LNA with source inductive
degeneration. Single ended LNAs have smaller noise figures
compared to differential LNAs, they also eliminate the need
for using a balun transformer to interface the antenna to a
differential LNA input, so it consumes less power. The
transistor M 1 and M 2 are same size RF transistors, giving a
better noise isolation. The common-gate transistor
M 2 increases the reverse isolation of the LNA, and improves
the stability of the circuit by minimizing the miller effect of
the parasitic gate-drain capacitance of M 1 (Cgd 1 ) by presenting

I. INTRODUCTION
The use of Radio Frequency in medical applications has
brought numerous advantages such as increased patient safety,
comfort and mobility, for these reasons highly reliable and
ultra-low power consumption implanted wireless devices are
mostly preferred in medical industry [1]. In direct conversion
receivers, the antenna receives the signal from the implanted
device, and the LNA amplifies the weak signal. The mixer
then uses the local oscillator signal to demodulate the input
which is then down converted to the base band.

a low impedance node at the drain of M 1 [4].

The LNAs presented in this paper are intended to be used

as the first stage of a Medical Implant Communications
Service (MICS) receiver to provide enough gain to overcome
the noise of subsequent stages while adding as little noise as
possible to the incoming signal. Besides, it should amplify
large signals without distortion, offer enough dynamic range
and provide input and output impedance matching [2]. LNA
design involves tradeoffs between linearity, input matching,
noise figure, and power dissipation.

Vdd
(W L)1 = 350 0.5
(W L)2 = 350 0.5
Ls = 1.6 nH
Lg = 80 nH
Ld = 14 nH
C d = 12 pF
C ex = 160 fF
VBias = 0.5V

The paper is organized as follows: Section II includes the

circuit analysis and design details of the single ended cascode
LNA with inductive source degeneration. Section III
introduces the design of the folded cascode (FC) LNA.
Section IV shows the design and the advantages of an LNA
using current reuse technique (CRT). The simulation results
and discussion pertaining to S-parameters, linearity and noise
are presented in section V. Conclusions are given in section
VI.

RFout

M2

Vbias

RFin

Ld

Cd

Lg
M1
Cex

Ls
Gnd
Figure 1. Cascode LNA with inductive source degeneration

1-4244-1342-7/07/$25.00 2007 IEEE

76

C. Noise Figure
A major advantage of the CS-LNA with inductive source
degeneration is that the resistive impedance is noiseless,
unlike other topologies where a noisy resistor is added in the
signal path to create a 50 termination impedance. The noise
in CS-LNA comprises three factors: channel noise, gate noise,
and correlated noise. The noise figure of the cascode LNA
with inductive degeneration can be computed as [3]:

The parallel LC-tank circuit Ld , Cd is chosen such that the

resonance frequency is at the RF signal of 403MHz and to
provide a DC path for the bias current of both M 1 and M 2 .
When the LC-tank resonates, it functions as an impedance
with a value Rb = (1 + Q 2 ) Rind . The operation of the LC-tank is
limited by the quality factor (Q) of the tank circuit [5]. A little
current will be lost by the LC-tank for parasitic resistance
( Rind ) of the on-chip spiral inductor and a loss factor ( ) is
defined to calculate the loss current.

1 gm 2
Rb 2 + 1 gm 2

F = 1+

where = 1 + 2 c QCS

A. Input Matching
The input impedance of the common-source (CS) LNA
can be written as [5]:
g L
1
+ m1 s
j Cgs1 Cgs1

and c is the correlation factor representing the correlation

between the gate noise and the drain noise. Note that the
miller capacitance Cgd has been neglected in the interest of
simplicity. We observe that (6) contains terms which are
proportional to QCS as well as inversely proportional to QCS .
Hence, there exists an optimum quality factor that minimizes
the noise figure. Also, the cutoff frequency T of the device
should be high, the channel length should be minimized.

(2)

where Cgs1 and gm1 are respectively the parasitic gate-to-source

capacitance and the transconduactance of M 1 . The input
matching at the resonance frequency (0 ) can be achieved by
setting the real part of (1) to the source impedance ( Rs ) and the
imaginary part to zero. The matching conditions are:
Rs =

Ls gm1
1
, Lg + Ls = 2
0 Cgs1
Cgs1

D. Linearity
An important metric of linearity for LNA design is the
input third order modulation point ( IIP3 ) of the circuit.
The IIP3 of the circuit in Fig.1, assuming input matching can
be written as [6]:

(3)

The input impedance of the CS-LNA takes the form of a series

resonance and the associated effective quality factor of the
input
circuit
assuming
input
matching
is
QCS = (1 20 Cgs Rs ) > 1 , the input matching can be adjusted by

IIP3 [dBm] = IIP3,in [dBm] 20log10 (1 0 Cgs Rs )

inductor Lg makes the use of a smaller Lg feasible, and eases

input matching.
B. Gain
The effective transconductance of the CS-LNA stage
neglecting the gate resistance is:
gm1

0 Cgs1 ( Rs + T Ls )

T
0 Rs (1 + T Ls Rs )

III. THE PROPOSED FOLDED CASCODE LNA

The power consumption is an important issue for radio
frequency integrated circuit design. A FC-LNA structure is
makes a low-voltage operation possible. The commongate M 2 of the cascode LNA is folded to another biasing path.
This FC topology gives the chance of increasing the effective
bias voltage. As the effective bias voltage increases, gm1 of
M 1 increases, consequently increasing the cutoff frequency,
which results in an overall reduction in the noise figure, (6).

(4)

It can be observed that the effective transconductance is

independent of the gm1 of the device, and is dependent on the
CMOS process parameters through the transition
frequency T . Using (4), we can calculate the following
expressions for the voltage gain assuming input matching.
A = (

T RL
)
0 2 Rs

(7)

The first term in (7) is the intrinsic IIP3 of the device, and
arises from the fact that short channel CMOS transistors
exhibit velocity saturation, which gradually linearizes the ideal
quadratic drain current equation of the long channel transistor.
The second term results from the extra voltage boost across
the Cgs due to the series tank.

tuning Lg and Ls . Adding a capacitor Cex in parallel with the

Gm ,CS = gm1QCS =

(6)

g
2 2
+
(1 + QCS2 ) ,  m , gd 0 is
5
5
gd 0
the zero-bias drain source conductance, and are the bias
dependent coefficients of channel thermal noise and gate noise
respectively, Rl represents the series resistance of the
inductor Lg , Rg is the gate resistance of the NMOS device,

(1)

The larger transconductance of M 2 ( gm 2 ) can be chosen to

minimize the lossy current.

Zin = j ( Lg + Ls ) +

Rl Rg 0
+
+
( )
Rs Rs QCS T

The circuit topology of a FC-LNA design is shown in

Fig.2. where the CS and CG stages are folded into two paths.
At DC the supply voltage needs to bias only one transistor in
each path, which mean that the minimum supply voltage

(5)

where RL is the load resistance of the LNA.

77

Fig.3 (b). The total transconductance is gmt = gm1 + gm 2 .

Vdd

Ld

Cd

Vbias
Lg

RFin

With (W L )1 = (W L )2 = (1 2) W L , the input capacitance

(W L)1 = 350 0.5 Ls = 1.6 nH

(W L)2 = 350 0.5 Lg = 80 nH
Ld = 14 nH Cd = 12 pF
L0 = 14 nH C 0 = 12 pF
C ex = 160 fF VBias = 0.5V

M1

is Cgs1 + Cgs 2 which is nearly equal to Cgs in Fig.3 (a).

Fig.4 shows the LNA topology. An inverter stage
( M 1 and M 2 ) with inductive source degeneration is used to
provide more design parameters to achieve a high gain and
low NF at the same time. This configuration provides a high
transconductance necessary to achieve high gain, and low
reverse gain necessary to provide sufficient isolation and to
simplify input and output port matching. The transistor M 3 is
folded to another biasing path. The NF of the LNA neglecting
the gate-drain capacitance can be shown to be [7]:

M2

Cex

RFout

Ls

C0

Gnd

L0

F 1 + (8 2 Cgs2 Rs ) 3gm1

Gnd

Figure 2. Folded cascode LNA schematic

This LNA in Fig.4 is capable of achieving the desired NF

specification where a large gm1 is typically needed to reduce
the noise figure.

needs only one threshold voltage (Vth ) to drive the transistor.

At AC the RF signal is fed from the CS stage into the CG
stage; and actually works as a traditional cascode LNA at the
operation frequency. The ( Ld , Cd ) tank is chosen such that the
resonance frequency is at the RF signal frequency 403MHz.
At this frequency it provides a high impedance branch to force
the RF signal into the source of M 2 . In order for this
operation to be true it is required that the impedance of the
parallel LC-tank at resonance is much higher than the input
resistance seen at the source of M 2 . It can be shown that this
is satisfied by choosing the transconductance of M 2 ,
gm 2  1 Rd Qd2 where Rd and Qd are the resistance and quality
factor associated with the inductor Ld . The tuned
tank L0 , C0 circuit acts as a filter to ensure that the
specifications for center frequency, bandwidth and gain are
met.

V. SIMULATION RESULTS AND DISCUSSION

The three designed LNAs are simulated using a
0.13 m CMOS process in the Cadence design environment.
The input signal power ranged from -40 to 0dBm. The Sparameter are used to describe the behavior of a system
working at high frequencies, where S11 represents input
reflection, S22 output reflection, S21 direct gain and S12 is the
reverse gain.
The software ASITIC [8] is then used to generate the
equivalent circuit model for spiral inductors to be included in
the simulation. These parasitic effects must be taken into
account since they have a significant effect on the overall
noise and the optimum device geometry and power.
Fig.5 shows the power gain ( S21 ) simulation results of the
three LNAs. The maximum gain is obtained from the LNA
using CRT, where a high transconductance is achieved. This
large gain gives the room to reduce the effect of noise from
the following stages. These three LNA circuits show a good
S11 (less than 25dBm ), S22 (less than 15dBm ), and input
output isolation. Furthermore, these amplifiers have a good
linearity.
V

IV. THE PROPOSED LNA USING CRT

Unlike traditional CMOS LNAs that trade off the noise
figure for input matching, the goal here is to achieve a high
transconductance gm and a cutoff frequency T = gm Cgs
necessary to achieve a high gain with less current. The current
reuse technique (CRT) topology presented in this paper is able
to simultaneously achieve a good input matching and a low
noise figure. Fig.3 (a) shows a single NMOS device that has
the aspect ratioW L with a drain current I D . Fig.3 (b) shows
two NMOS devices in parallel with each device having an
aspect ratio (1 2)W L and drain current (1 2) I D . Thus, the
transconductance of the compound device in Fig.3 (b) is the
same as the transconductance of the device in Fig.3 (a). In
Fig.3(c), a PMOS device is substituted for device M 2 in
gm
M1

gm ID
ID

(1 / 2)ID

(a)

1/2(W/L)

dd

Cd

Ld

Vbias
RFin

Lg

M2

Cex

M1

(1/ 2)I D

(W L)1 = 500 0.13 Ls = 1 nH

(W L)2 = 500 0.13 Lg = 80 nH
(W L)3 = 600 0.13 Cex = 160 fF
Ld = 14 nH Cd = 12 pF
L0 = 12 nH C 0 = 12 pF

M3

M2

(1/ 2)I D

M1

W/L

(W/L) 2

(8)

gm1 + gm2

M2
1/2(W/L)

(b)

Ls

M1
(W/L)1

(c)

C0

L0

(1/ 2)I D

Gnd
Gnd
Figure 4. Degenerated current reuse LNA

Figure 3. Illustration of CRT technique

78

RFout

The FC-LNA as indicated in Fig.6 has the best linearity where

only one single transistor exists in each DC path which
increases the voltage swing and consequently improves the
circuit linearity. Another important figure of merit is the noise
figure; all three LNAs have a fairly low NF over a frequency
band as shown in Fig. 7. The total power dissipation for the
FC-LNA and CRT-LNA is the best due to the high
transconductance with a lower current. Table 1 summarizes
the simulation results obtained from the simulations.

3rd Order
1st Order

20

Output Power (dBm)

VI.

40

CONCLUSIONS

This paper proposes a low power design techniques for

LNA with a 1-V power supply in a 0.13 m CMOS process.
These structures are based on CS-LNA with inductive source
degeneration. The proposed LC-folded cascode topology
maintains the advantages of the cascode LNA structure while
making it feasible for the input amplifier of the cascode
structure to work at a high effective voltage with a less
current. This improves the circuit linearity, increases the input
transconductance and consequently the power gain. The CRTLNA is designed to increase the transconductance of the input
stage at the same bias current to be able to achieve a high gain
with a low noise figure. As a tradeoff an additional LC-tank
circuit is used for both FC-LNA and CRT-LNA which
certainly is a major area contributor. Simulation results show
that the three LNAs designed in this work meet the MICS
requirements.
TABLE I.
Parameter

FC-LNA

16.1

18.2

23.7

Isolation (dB)

-60.2

-62.03

-40.2

S11 (dB )

-21.8

-31.78

-41.2

S 22 (dB)

-13.4

-19.5

-18.35

NF (dB )

0.65

0.60

0.62

IIP3 (dBm)

-9.38

-8.0

-10.4

1dB (dBm)

-19.02

-17.71

-25.5

Power cons.(mW)

1.0

0.8

0.8

16

S21 (dB)

NF (dB)

10
8
6

2
0
0

200

400

600

800

1000

Frequency (MHz)

Figure 7. The noise figure of the three LNAs

REFERENCES
[1]

[2]

[5]

[6]

-10

[7]

-15
-20
-25
800

-5

600

-5

NF (Cascode)

400

-10

12

S21 (Cascode)

200

-15

NF (CRT)
NF (FC)

14

[4]

-20

18

S21(CRT)
S21 (FC)

10

-25

Figure 6. Input third-order intercept point of the FC-LNA

25

15

IIP3 = - 8.0 dBm

Input Power (dBm)

[3]

20

-40

-30

CRT-LNA

Gain (dB)

-20

-60

LNAS SIMULATION RESULTS SUMMARY

Cascode LNA

[8]

1000

Frequency (MHz)

Figure 5. The gain (S21) of the three LNAs

79

Huseyin S., Ahmet Sula, Zheng Wang, Numan S., E. Arvas, MICS
transceivers: regulatory, standards and applications, IEEE Southeast
Con.2005, pp. 179-182, April 2005.
R.L. Moreno and E.C. Rodrigues, CMOS LNA for wireless
biomedical telemetry, IEE Proc. Circuits Devices and Systems, vol.
152, pp.401-406, October 2005.
Derek K. Shaeffer, and Thomas H. Lee, A 1.5-V, 1.5-GHz CMOS
Low Noise Amplifier,IEEE Journal of Solid-State Circuits, vol.32,
no.32, pp. 745-759, May 1997.
Huseyin S., Ahmet Sula, Zheng Wang, Numan S., E. Arvas, A 1-V
UHF low noise amplifier for ultralow-power applications, proc. IEEE
2006 ISCAS06, pp. 4495-4498, May 2006.
David J. Allstot, Xiaoyong Li, and Sudip Shekhar, Design
considerations for CMOS low-noise amplifiers, symp. IEEE on RFIC,
pp.97-100, June 2004
Paul Leroux, Johan Janessens, and Michiel Stayaert, A 0.8-dB NF
ESD-protected 9-mW CMOS LNA operating at 1.23 GHz, IEEE
Journal of Solid-State Circuits, vol.37, no.6, pp. 760-765, June 2002.
Andrew N. Karanicolas, A 2.7-V 900-MHz CMOS LNA and mixer,
IEEE Journal of Solid-State Circuits, vol.31, no.12, pp. 1939-1944,
December 1996.
A. M. Niknejad and R. G. Meyer, Analsis and optimization of
monolithic inductors and transformers for RF ICs, Proc. IEEE 1997
CICC, pp.375-378, May 1997.