MOS Diff LNA - courses.e-ce.uth.gr · Sheet 2 of 14 Rs = 50Ω Lg M1 Using Agilent CMOS14 process...

Sheet

1 of 14

MOS Differential LNA Design Tutorial J P Silver

E-mail: [email protected]

1 ABSTRACT

This tutorial describes the theory and design on a MOS Differential Low noise amplifier using source de-generation. A worked example is given together with the associated Agilent ADS simulation circuits and plots.

2 INTRODUCTION The MOS LNA Design tutorial describes the design of a single stage LNA using source degeneration technique to provide a good noise match. A cascade output stage was added to the source degenerated stage provide improved gain & reverse isolation. This tutorial describes the design of a differential LNA using the same design specification and device models. There are several advantages in using a differential de-sign. Firstly, the virtual ground formed at the ‘tail’ re-moves the sensitivity to parasitic ground inductances, which makes the real part of the input impedance purely controlled by the source degeneration inductance (Ls). Secondly the differential amplification of the signal en-sures attenuation of the common mode signal, in most systems this common mode signal will be noise! Thirdly, the use of Gilbert mixers and image rejection schemes require to be fed from a differential source.

3 LNA SINGLE-STAGE DESIGN The design equations are the same as for the single stage LNA design and are summarized below:

[ ]gs

msCGSLS

gss

gs

ms

C.gL

Ra WhereX-Xj Ra Rg Rin

as written-re be Can ωC

1ωLj C.gL

Rg Rin

=++=

⎟⎟⎠

⎞⎜⎜⎝

⎛−++=

Therefore, the impedance of the MOSFET without feed-back is:

CGSCGS jX RinjX Rg Rin −=→−=

Adding series feedback adds the following term to the original input impedance:

jX Ra LS+

Additionally, another inductor is added in series with the gate Lg that is selected to resonate with the Cgs Capacitor. What we are trying to achieve is:

0ωC

1ωLj

ie Cgs out cancels it frequency resonant the at that so designed is Lg

ohms. 50say bemay Rin WhereC.gL

Rin

gss

gs

ms

=⎟⎟⎠

⎞⎜⎜⎝

⎛−

=

In most LNA designs the value of Ls is picked and the values of gm and Cgs are calculated to give the required Rin.

4 DESIGN EXAMPLE The aim of this example is to design step-by-step a nar-row band LNA (Low noise amplifier) to work over the Bluetooth frequency band. A summary of the required specification for the LNA is given in Table 1. Parameter Specification Units Frequency 2.45 to 2.85 GHz Noise Figure <3.5 dB Voltage Gain >20 dB Power Gain >10 dB Power consumption <100 mW Source/load impedance 50 ohms Load Capacitance 0.4 pF

Table 1 Required specification for the Bluetooth front end LNA.

For this design we will be using the Agilent CMOS14 0.5um process that allows a minimum gate length of 0.6um. The schematic of the single ended LNA (half of the final differential LNA design) is shown in Figure 1.

Sheet

2 of 14

Rs = 50Ω

Lg

M1 Using Agilent

CMOS14 process

Vbias

Ls

Lbias

Vcc +1.5V

Figure 1 Initial single-stage LNA schematic

4.1 STARTING VALUE OF DE-GENERATION INDUCTOR LS. The value of this inductor is fairly arbitrary but is ulti-mately limited on the maximum size of inductance al-lowed by the technology, which is typically about 10nH (anything bigger would probably be too big to be put on the chip). For this example we will pick a value of 1.0nH. We now find the cut-off frequency defined as:

16GHz)(~rad/sec 1E 0.5E

50 LsRs ω 11

9-T ====Cgsgm

4.2 OPTIMAL Q OF INDUCTOR Optimal Q is given by:

.γδ.α5

p Where

p11QL

2

=

+=

The parameters for p are dependant on the CMOS tech-nology but typically

0.9) be to (take 1-0.8 be to assumed is α4)(normally of value the times 3-2 to set is δ

2)(normally 3 - 2 between set is γ

γ

2.67 0.162

11QL

0.162 45

)9.0(4 p2

=+=

==.

.

4.3 EVALUATION OF LG

7.52nH 1E-1.665E2.67.50 Lg

rad/sec 1.665E 2.65E2

frequency centreω Where

Ls-ω.RsQ Lg

9-10

109

o

o

L

==

=

==

=

π.

4.4 FIND CGS (GATE-SOURCE CAPACITANCE)

( )

( ) ( )pF45.0

E5.07.52E1.665E

1Cgs

LsLgω1Cgs

99210

2o

=+

=

+=

−−

4.5 FIND W

330umW

330 .0.63.419E

0.45.23 W

pF/um3.419E 1.01E

8.854E x 3.9 ToxεoxCox

F/cm8.854E space free for constant dielectric εoand 3.9 silicon for constant dielectric εs

Where

εo .εox εox

m1.01E Tox m;0.6E Lmin

Cox.LCgs

23 W

ie Wget to rearrange Cox.W.L32 Cgs

3-

23-8-

14-

14-

8-6-

min

min

=

==

===

==

==

=

==

=

=

Sheet

3 of 14

4.6 CALCULATE GM

0.045A/V .0.45E1Egm

.Cgsωgm

12-11

T

==

=

4.7 V EFFECTIVE

gate the toV 16.1 0.5)(0.67

apply to need weTherefore,

0.67V VTWith

0.5V 164.330

45000.0.6Veff

:gives units Converting433cm/V mobility deviceu

.Cox.Wugm.L

)V(VgsVeff

n

n

minT

=+

=

==

==

=−=

4.8 BIAS CURRENT ID

mA11.0.5045.0.21 gm.V.

21 Id eff ===

(9) Estimated Optimum Noise Figure

( )

[ ]( )

3.26dB )10log(2.12 2.12Fopt

162.01162.04.0162.05E

1.665E0.941

Fopt4.0c Take

1cωω

α2γ1Fopt

10

10

T

o

===

+++⎟⎟⎠

⎞⎜⎜⎝

⎛+

=

=

+++⎟⎟⎠

⎞⎜⎜⎝

⎛+= ppp

To complete the design another N-type is added as the cascode stage with the same W/L ratio as the LNA device. The cascode stage is connected at a iode ie the gate is DC connected to the drain to en-sure the device is in saturation. Finally it is as-sumed that on-chip spiral inductors will be used in the design, which will have a typical loaded Q of 6.

5 DIFFERENTIAL LNA DESIGN The basic circuit of the differential LNA is shown in Figure 2. As can be seen each half of the differ-

ential amplifier is in fact the single LNA designed earlier in this tutorial (ie with a 50ohm input im-pedance set by making gm = 20mS), with the de-generating inductors (Ls) connected together at the ‘virtual earth’. At this point the a current source is connected to the negative supply set to give twice the current flowing down one of the LNA sections. The ADS schematic showing the differential LNA is shown in Figure 3. Note to supply a differential signal to each LNA input, an ‘ideal’ balun (bal-anced to unbalanced) transformer has been used (We could also use two AC sources each set to 0.5V and opposite polarity). In addition another balun is used on the amplifier output to re-combine the signal to allow the voltage gain to be simulated.

Rs = 50Ω

Lg

All devices use Agilent

CMOS14 process

Vout1

Ls

Lbias

Vcc +2.5V

Vin1

Ls Rs = 50Ω

Lg

Vin2

Vout2

Lbias

Vee -2.5V

‘Tail’ current bias

M5 ‘Virtual’ Ground

11mA

M1

22mA

M2

M3 M4

Figure 2 Schematic of the differential amplifier LNA using two single stage LNA’s from the previous tutorials. Circuit values are Lg = 7.52nH ; Ls = 1nH; M1,M2,M3 & M4 W/L = 330 (Lmin = 0.5um).

Calculation of Load Inductor:

10nH 0.4E

E5.2.21

C

o.21

L

L then 0.4pF of capacitor load a Assumimg

.LC21o

12-

2

9

out

2

LOAD

Load

Loadout

=⎟⎠⎞

⎜⎝⎛

=⎟⎠⎞

⎜⎝⎛

=

=

=

πfπ

πf

Sheet

4 of 14

vg

v2v1

Balun3PortCMP1

V_ACSRC6

Freq=freqVac=polar(1,0) V

INDQL4

Rdc=0.0 OhmMode=proportional to freqF=2400.0 MHzQ=6L=Lg nH

CC5C=10 pF

VARVAR4

Lg=5.5Load=7.5Ls=0.5

EqnVar

CC4C=0.4 pFC

C3C=0.4 pF

INDQL5

Rdc=0.0 OhmMode=proportional to freqF=2400.0 MHzQ=6L=Load nH

VARVAR3Ibias=22

EqnVar

RR2R=10k

MOSFET_NMOSMOSFET4

Width=W umLength=L umModel=cmosn

vee

MOSFET_NMOSMOSFET6


vee

INDQL6

Rdc=0.0 OhmMode=proportional to freqF=2400.0 MHzQ=6L=Ls nH

DCDC1

DC

BSIM3_Modelcmosn

Xl=-1e-7Xw=0Em=4.1e7Kt2=0.022Kt1=-0.11Uc1=-5.6e-11

Ub1=-7.61e-18Ua1=4.31e-9At=3.3e4Ute=-1.5Pvag=0.1945781Pscbe2=5e-10Pscbe1=2.541131e10Pdiblcb=-0.5Pdiblc2=9.723614e-4Pdiblc1=2.091364e-3Pclm=0.7319137Drout=0.0428851Dsub=0.751089Etab=2.603903e-3Eta0=0.1178659

Cit=0Cdscd=0Cdscb=0Cdsc=2.4e-4Nfactor=1.2410485Voff=-0.0850186B1=5e-6B0=1.648829e-6Pags=0.0968Ags=0.1450882Wketa=-5.792854e-3Lketa=-0.0143698Keta=3.997018e-3A0=0.9059229Vsat=1.174604e5

Prwb=-0.0586598Prwg=0.0182608Prdsw=-33.9337286Rdsw=1.28604e3Delta=0.01Uc=1.831708e-11Ub=1.582544e-18Ua=1e-12Dvt2=-0.1427458Dvt1=0.9107896Dvt0=6.5803089Nlx=5.28517e-8W0=1e-5K3b=1.252205K3=68.279056

Pk2=9.631217e-3K2=-0.0316751K1=0.825917Pvth0=8.691731e-3Vth0=0.6701079U0=433.6065339Xj=1.5e-7Vbm=-3.0Nch=1.7e17Dwb=1.238214e-8Dwg=-7.483283e-9Xpart=0.5Cgbo=2e-9Cgdo=2.79e-10Cgso=2.79e-10

Pbswg=0.99Mjswg=0.1Cjswg=2.2346e-10Pbsw=0.99Pb=0.99Mjsw=0.1Cjsw=4.437149e-10Mj=0.7549569Cj=5.067009e-4Tox=1.01e-8Tnom=27Wwl=0Wwn=1Ww=0Wln=1

Wl=0Wint=2.277646e-7Lwl=0Lwn=1Lw=0Lln=1Ll=0Lint=1.097132e-7Js=0Rsh=2.8Capmod=2Mobmod=1Version=3.1Idsmod=8NMOS=yes

RR1R=10k

VARVAR1

W=330L=0.6

EqnVar

V_DCSRC1Vdc=2.5 V

I_ProbeID

INDQL1


MOSFET_NMOSMOSFET1


vee

CC2C=10 pF

MOSFET_NMOSMOSFET3


vee

INDQL3


INDQL7


ACAC1

Step=Stop=3.0 GHzStart=2.0 GHz

AC

I_DCSRC5Idc=Ibias mA

V_DCSRC4Vdc=-2.5 V

Figure 3 ADS schematic of the Common-source Differential LNA. The differential input voltage is supplied vai the balun transformer (CMP1). An ideal current source has been added at this stage on the ‘tail’ set to 22mA so that 11mA will flow down each ‘arm’ of the LNA. The 0.4pF load capacitors (representing a typical load of say a gilbert mixer) have been added to each of the two output nodes V1 & V2.

Sheet

5 of 14

vg

v2v1

CC5C=22 pF

CC2C=22 pF

VARVAR3Ibias=22

EqnVar

DCDC1

DC

VARVAR4


EqnVar

VARVAR1

W2=330W1=330L=0.6

EqnVarS_Param

SP1

Step=Stop=3 GHzStart=2 GHz

S-PARAMETERS

TermTerm1

Z=50 OhmNum=1

Balun3PortCMP1

V_DCSRC1Vdc=2.5 V

RR2R=5 kOhm

RR1R=5 kOhm

MOSFET_NMOSMOSFET6

Width=W1 umLength=L umModel=cmosn

vee

MOSFET_NMOSMOSFET4


vee

MOSFET_NMOSMOSFET3


vee

MOSFET_NMOSMOSFET1


vee

TermTerm2

Z=50 OhmNum=2Term

Term3

Z=50 OhmNum=3

DC_BlockDC_Block2

DC_BlockDC_Block1

INDQL4


INDQL5


INDQL6


BSIM3_Modelcmosn








I_ProbeID

INDQL1


INDQL3


INDQL7



V_DCSRC4Vdc=-2.5 V

Figure 4 ADS schematic of the Common-source Differential LNA. The small signal input power is supplied via the balun transformer (CMP1). An ideal current source has been added at this stage on the ‘tail’ set to 22mA so that 11mA will flow down each ‘arm’ of the LNA. The two output nodes V1 & V2 are now terminated in 50-ohm loads – Note the DC blocks to ensure that the DC bias condition, is not upset by the loads.

Sheet

6 of 14

5.1 ESTIMATION OF DIFFERENTIAL GAIN

32dB20log(41)(dB)A

412.5E x 2π x 0.4E

6 x 0.043 Co.2πo.2gm.Q

A

0.043 E11 0.6300.170E x 2 gm

I.LW2.Kp. gm

LNA

912-LOAD

LNA

36-

DS

==

===

==

==

−x

This gain is for one C-S stage and as the output are not combined the gain of the differential amplifier will be half that for a single stage ie 3dB less. This will make the predicted gain of our example 29dB. Note to centre the gain response the load inductors were reduced to 7nH.

m1 freq=2.493GHz dB(AC.vout1)=27.176

2.2 2.4 2.6 2.8 2.0 3.0

16 18 20 22 24 26

14

28

freq, GHz

dB(AC.vout1) m1

DC.ID.i 11.00mA

Figure 5 Simulation result of the ADS schematic shown in Figure 3. The predicted voltage gain is 29dB, which is pretty close to the simulated re-sult and much greater than the specification of 20dB. Note to centre the gain response, the load inductors were reduced from 10nH to 7nH.

5.2 S-PARAMETERS The differential voltage gain is predicted to be ~29dB exceeding our specification for voltage gain, but power gain requirement has to met too. To check the power gain ie S21 we need to configure the ADS schematic of the LNA with an S-Parameter simulator. This is show in Figure 4 with the resulting simulation result shown in Figure 6.

DC.ID.i11.00mA

m1 freq=2.513GHzdB(S(2,1))=8.315

2.2 2.4 2.6 2.8 2.0 3.0

6.5

7.0

7.5

8.0

8.5

6.0

9.0

freq, GHz

dB(S(2,1))

m1

m2 freq=2.503GHznf(2)=2.877

2.2 2.4 2.6 2.8 2.0 3.0

2.7

2.8

2.9

3.0

3.1

3.2

2.6

3.3

freq, GHz nf(2)

m2

Figure 6 Simulation of power gain (S21) and noise figure of the LNA using the setup shown in Figure 4.

The simulation plot of Figure 6, shows the classic gain/noise trade-off of LNA design. To peak the gain at 2.5GHz we need to increase the gate inductance (Lg), to peak the noise we need to decrease the noise figure. The noise specification for this LNA is <3.5dB so there is a little margin to allow for gain peaking, but the power gain is too low at 9dB. There are a number of ways to increase the gain: (1) Increase W/L ratio of cascode. (2) Increase current. (3) Reduce Ls. (4) Add another C-S stage to each output node. The effect of optimizing Lgate for best power gain is shown in Figure 7, however although the noise figure is within the specification the power gain is still short of 10dB. By adjusting the current, W/L ratios and device inductances (ie Lg & Ls) it was possible to achieve the required design goals for gain and noise at the expense of power consump-tion.

Sheet

7 of 14

DC.ID.i 11.00mA

m1 freq=2.513GHz dB(S(2,1))=9.271

2.2 2.4 2.6 2.8 2.0 3.0

8.0

8.5

9.0

7.5

9.5

freq, GHz

dB(S(2,1)) m1

m2freq=2.503GHz nf(2)=3.135

2.2 2.4 2.6 2.8 2.0 3.0

3.2

3.3

3.4

3.1

3.5

freq, GHz

nf(2)

m2

Figure 7 ADS simulation of the circuit shown in Figure 4 with each Lgate increased to 7nH to cen-ter the gain response at 2.5GHz. Note that the noise figure has now degraded to 3.1dB. Further optimization is required to increase S21 to > 10dB.

DC.ID.i 20.00mA

m1 freq=2.513GHz dB(S(2,1))=10.883

2.2 2.4 2.6 2.8 2.0 3.0

10.0 10.2 10.4 10.6 10.8

9.8

11.0

freq, GHz

dB(S(2,1)) m1

m2freq=2.503GHz nf(2)=2.397

2.2 2.4 2.6 2.8 2.0 3.0

2.40 2.45 2.50 2.55

2.35

2.60

freq, GHz

nf(2)

m2

Figure 8 ADS simulation of the circuit shown in Figure 4 with Lg decreased to 4.5nH, Ls reduced to 0.35nH, Cascode width increased to 660um, Input device width increased to 530um and the current through each arm increased to 20mA .

Although the specification for power gain is now met, a 1.8dB margin might be unacceptable when taking into account process and temperature varia-tions. Therefore, the final design gain is increased by adding simple C-S stages with inductive loads and decoupled on the output by small value capaci-tor. This output stage is DC coupled to the output of the first ‘cascode’ stage so that is receives a cor-rect bias to be in saturation. The increased gain will greatly improve the noise figure of the receiver as the noise figure of the second stage (most likely the mixer) will be reduced by ~1/gain_LNA. The modified circuit and increased gain responses are shown in Figure 9 and Figure 10.

DC.ID.i11.00mA

m1freq=2.490GHz dB(S(2,1))=24.145

1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.41.4 3.6

16

18

20

22

24

14

26

freq, GHz

dB(S(2,1))

m1

m2freq=2.498GHz nf(2)=2.936

1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.41.4 3.6

3.0

3.5

4.0

2.5

4.5

freq, GHz

nf(2)

m2

Figure 9 Gain & noise plots from the differential LNA with added C-S stage on the output (with a tuned inductive load).

Sheet

8 of 14

v2

vg

v1

INDQL9

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6L=Load

MOSFET_NMOSMOSFET7

Width=330 umLength=L Model=cmos

CC3C=0.4 pF

Term Term1 Z=50 Ohm Num=1

TermTerm2

Z=50 OhmNum=2

S_ParamSP1


S-PARAMETERS

CC6C=10 pF

INDQ L8

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6 L=1 nH

MOSFET_NMOS MOSFET4

Width=W um Length=L Model=cmos

vee

VARVAR4


EqnVar

Balun3PorCMP1

INDQL4

Rdc=0.0 Ohm Mode=proportional to F=2400.0 MHz Q=6L=Lg nH C

C5C=10 pF

INDQL5

Rdc=0.0 Ohm Mode=proportional to F=2400.0 MHz Q=6L=Load

VARVAR3Ibias=22

EqnVar

R R2 R=10k

MOSFET_NMOS MOSFET6


vee

INDQL6

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6L=Ls nH

DC DC1

DC

BSIM3_Modecmosn

Xl=-1e-7 Xw=0 Em=4.1e7Kt2=0.022Kt1=-0.11Uc1=-5.6e-

Ub1=-7.61e-Ua1=4.31e-At=3.3e4 Ute=-1.5 Pvag=0.194578Pscbe2=5e-Pscbe1=2.541131ePdiblcb=-Pdiblc2=9.723614ePdiblc1=2.091364ePclm=0.731913Drout=0.042885Dsub=0.75108Etab=2.603903e-Eta0=0.117865

Cit=0Cdscd=0Cdscb=0Cdsc=2.4e-4Nfactor=1.241048Voff=-B1=5e-6B0=1.648829e-Pags=0.096Ags=0.145088Wketa=-5.792854e-Lketa=-Keta=3.997018e-A0=0.905922Vsat=1.174604e

Prwb=-Prwg=0.018260Prdsw=-Rdsw=1.28604eDelta=0.01Uc=1.831708e-Ub=1.582544e-Ua=1e-12Dvt2=-Dvt1=0.910789Dvt0=6.580308Nlx=5.28517e-W0=1e-5K3b=1.25220K3=68.27905

Pk2=9.631217e-K2=-K1=0.82591Pvth0=8.691731e-Vth0=0.670107U0=433.606533Xj=1.5e-7Vbm=-3.0Nch=1.7e1Dwb=1.238214e-Dwg=-7.483283e-Xpart=0.5Cgbo=2e-Cgdo=2.79e-Cgso=2.79e-

Pbswg=0.99Mjswg=0.1 Cjswg=2.2346e-Pbsw=0.99 Pb=0.99 Mjsw=0.1 Cjsw=4.437149e-Mj=0.754956Cj=5.067009e-Tox=1.01e-8 Tnom=27 Wwl=0Wwn=1 Ww=0Wln=1

Wl=0 Wint=2.277646e-Lwl=0 Lwn=1 Lw=0 Lln=1 Ll=0 Lint=1.097132e-Js=0 Rsh=2.8 Capmod=Mobmod=1 Version=3.1Idsmod=8 NMOS=yes

R R1 R=10k

VARVAR1

W=330L=0.6

EqnVar

V_DCSRC1Vdc=2.5 V

I_Probe ID

INDQ L1

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6 L=Lg nH

MOSFET_NMOS MOSFET1


vee

C C2 C=10 pF

MOSFET_NMOSMOSFET3


vee

INDQL3

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6L=Load

INDQL7

Rdc=0.0 OhmMode=proportional to F=2400.0 MHzQ=6L=Ls nH


V_DCSRC4Vdc=-2.5 V

Figure 10 ADS schematic of the differential LNA with a C-S stage added to the cascode output. Note for clar-ity only one extra C-S stage has been added, normally another stage would be added to M1. To complete the design a current mirror (and associated bias network) will replace the ideal 11mA current source (SRC5).

6 LINEARITY This final section deals with estimating the 1dB gain compression point and IM3 (3rd order inter-modulation products). If an amplifier is driven hard enough the output power will begin to roll off re-sulting in a drop of gain known as gain compres-sion. The measurement of gain compression, is given by the 1dB gain compression point. This pa-rameter in another measure of the linearity of a de-vice and is defined as the input power that causes a 1dB drop in the linear gain due to device saturation. An example of the 1dB compression point is shown in Figure 11.

We can get a rough estimate of the gain compres-sion of the LNA a non-linear expression of the in-put and output parameters can be expanded using Taylor’s theorem. This results in the following equations for 1dB gain compression point and IM3 [3].

Sheet

9 of 14

1dB Compression Point 20dBm

0

5

10

15

20

25

30

-10 -5 0 5 10

Pin (dBm)

Pout (dBm)

1dB gain change

Saturated Output Power 21dBm

Figure 11 Compression point plot. The red curve is the gain compression characteristic of a 15dB gain amplifier. The parallel black line is 1dB be-low and represents a gain of 14dB. The input power where the amplifier gain crosses the 1dB down line is know as the 1dB input compression point.

od.V1eff Whereeff

sat2V satE

by given strength field saturation Velocity satE

and Vt- Vgs overdrive Voltage odV

With

.LsatEodV

2odV

.ox.CvsatWDSATI

ϑ+==

=

==

+=

oµµµ

(1) - .2

.11.4

.11.1

.29.0~2

1 ⎟⎟⎠

⎞⎜⎜⎝

⎛+⎟

⎟⎠

⎞⎜⎜⎝

⎛+

LV

Vµ

LV

VµV

RsµLVP

sat

od

sat

odod

satdB

to convert to dBm = 10 log (1000*P1dB)

Vsat.L2oµ1µ

(2) - .L2V

.V1µ1

.L4V

.V1µ1V

.Rs1µ.LV

38~

2

sat

od

sat

od

odsat

IIP3

ϑ+≅

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛+

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛+P

Example Running the simulation shown in Figure 10, the values of Vgs = 1.26V. Using the spice model data for the Agilent CMOS14 0.5um we have: L = 0.6um, µ0 = 433 cm^2/(V*s), θ = 0.1, Rs = 50 ohms, VT = 0.67V Vsat = 1.73E5 m/s First convert numbers to metre format: µ0 = 433 cm^2/(V*s), = 0.433 m^2/(V*s L = 0.6um, = 0.6E-6m.

0.59V 0.67-1.26 TV-gs VodV ===

453 .0.6EE2x0.1x1.730.4331

65 =+≅ −µ

Feeding these values into equations 1 & 2 yields P1dB = 9.39dBm and IIM3 = 19.02dBm. The equations 1 & 2 were entered into a spread-sheet, along with a range of Vod from 0.01 to 5V. The table of results is shown in Table 2.

Vod (V) Input 1dB Comp IIMP3 (dBm) (dBm)

0.01 -10.05 -0.41 0.5 8.42 18.05 0.59 9.39 19.02

1 12.77 22.40 1.5 15.74 25.37 2 18.09 27.72

2.5 20.07 29.70 3 21.80 31.43

3.5 23.34 32.97 4 24.73 34.37

4.5 26.01 35.64 5 27.19 36.82

Table 2 Calculated 1dB compression point & 3rd order intercept point.

Sheet

10 of 14

The following two graphs show the plotted data of the data shown in Table 2. 1dB compression point is shown in & 3rd order in-tercept point is shown in Figure 12 and Figure 13 respectively.

Input referred 1dB compression point vs Vod

-15

-10

-5

0

5

10

15

20

25

30

0 1 2 3 4 5

Vod = Vgs-Vt (Volts)

Inpu

t 1dB

com

p (d

Bm

)

Figure 12 Referred to input 1dB compression point plot using the CMOS 14 Spice data

Input referred 3rd Order Intercept Point vs Vod

-5

0

5

10

15

20

25

30

35

40

0 1 2 3 4 5

Vod = Vgs-Vt (Volts)

Inpu

t 1dB

com

p (d

Bm

)

Figure 13 Referred to 3rd Order Intercept point (IIMP3) plot using the CMOS 14 Spice data

Figure 14 shows the ADS simulation setup to al-low the 1dB compression point to be simulated. The sub-circuit Diff_MOS_LNA is shown in Figure 15. The resulting plots are shown in Figure 16 (harmonic spec-trum) and Figure 17 (gain compression plot).

Vout

Diff_MOS_LNAX1

SweepPlanPlan1

Reverse=noSweepPlan=UseSweepPlan=Start=-40.0 Stop=10 Step=0.5 Lin=

SWEEP PLAN

TermTerm2

Z=50 OhmNum=2

P_1TonePORT1

Freq=RFfreqP=dbmtow(Pin)Z=50 OhmNum=1

VARVAR1

Pin=-5RFfreq=2.5 GHz

EqnVar

XDBHB2

GC_MaxInputPower=100GC_OutputPowerTol=1e-3GC_InputPowerTol=1e-3GC_OutputFreq=2.5 GHzGC_InputFreq=2.5 GHzGC_OutputPort=2GC_InputPort=1GC_XdB=1Order[1]=5Freq[1]=RFfreq

GAIN COMPRESSIONHarmonicBalanceHB1

SweepPlan="Plan1"SweepVar="Pin"Order[1]=5Freq[1]=RFfreq

HARMONIC BALANCE

Figure 14 ADS simulation setup to measure the 1dB compression point of the LNA. The sweep plan sets the input power sweep. The harmonic balance allows the LNA harmonics to be plotted and the gain compression block allows a pin vs pout plot to be generated.

Sheet

11 of 14

vg

v2v1

PortP1Num=1

CC3C=0.4 pF

PortP2Num=2

Balun3PortCMP2

Balun3PortCMP1

VARVAR4


EqnVar

INDQL4


CC5C=10 pF

CC4C=0.4 pF

INDQL5


VARVAR3Ibias=22

EqnVar

RR2R=10k

vee

MOSFET_NMOSMOSFET4


vee

MOSFET_NMOSMOSFET6


INDQL6


BSIM3_Modelcmosn








RR1R=10k

VARVAR1

W=330L=0.6

EqnVar

V_DCSRC1Vdc=2.5 V

I_ProbeID

INDQL1


vee

MOSFET_NMOSMOSFET1


CC2C=10 pF

vee

MOSFET_NMOSMOSFET3


INDQL3


INDQL7



V_DCSRC4Vdc=-2.5 V

Figure 15 Addition of Baluns on the Differential amplifier to allow gain compression simulation. Note the polarity of signals on the output balun.

Sheet

12 of 14

Eqn Gain=dBm(HB1.HB.Vout[1])-HB1.HB.Pin

Eqn linear=Gain[0]+HB1.HB.Pin

1 2 3 4 5 6 7 8 90 10

-300

-250

-200

-150

-100

-50

0

-350

50

freq, GHz

dBm

(HB2

.HB

.Vou

t)

Output Spectrum at 1 dB Gain Compression Point

Input and Output Powers at 1 dB Compression Point

inpwr[1]-2.271

outpwr[1]5.070

1dB Compression Characteristic of MOS LNA

Figure 16 Harmonic spectrum of the MOS LNA with a prediction of the 1dB compression point of +5.-7dBm (-4dBm input referred). The gain compression calculation agrees with the simulation value to within 1dB at 7.5dBm (Simula-tion value calculated at 6.5dBm). Finally, to increase the gain margin of the LNA another C-S stage could be added to the ’cascode’ output. This output stage is DC coupled to the out-put of the first ‘cascode’ stage so that is receives a correct bias to be in saturation. The increased gain will greatly improve the noise figure of the receiver as the noise figure of the second stage (most likely the mixer) will be reduced by ~1/gain_LNA.

Eqn compression=m1-m2

m1indep(m1)=plot_vs(dBm(HB1.HB.Vout[1]), HB1.HB.Pin)=5.179

-2.000

m2Pin=linear=6.341

-2.000

m1indep(m1)=plot_vs(dBm(HB1.HB.Vout[1]), HB1.HB.Pin)=5.179

-2.000

m2Pin=linear=6.341

-2.000

-30 -20 -10 0-40 10

-30

-20

-10

0

10

-40

20

HB1.HB.Pin

dBm

(HB1

.HB.

Vout

[1]) m1

Pin

linea

r

m2

compression-1.162

Figure 17 Gain compression plot of the MOS LNA. The markers m1 and m2 are adjusted to give a compression of 1dB, giving an output 1dB compression point of +6.3dBm. We could combine the outputs of the LNA by add-ing a current mirror to the cascode stages as shown in Figure 18. Note care has to be taken as the addi-tional current mirror will drop voltage and it may be necessary to raise the supply rail slightly. Alter-natively, the cascaded LNA could feed a second stage differential amplifier stage. This amplifier could then be loaded with a current mirror to sum the outputs.

Sheet

13 of 14

vout

v2v1

vg

V_DCSRC1Vdc=2.5 V

INDQL8


INDQL3


vee

MOSFET_NMOSMOSFET4

Width=W umLength=L umModel=cmosnvee

MOSFET_NMOSMOSFET3


ACAC1


AC

DCDC1

DCV_ACSRC6

Freq=freqVac=polar(1,0) V

vee

MOSFET_NMOSMOSFET8

Width=W umLength=L umModel=cmosn vee

MOSFET_NMOSMOSFET7


I_ProbeID

CC4C=0.4 pF

VARVAR4


EqnVar

Balun3PortCMP1

INDQL4


CC5C=10 pF

VARVAR3Ibias=22

EqnVar

RR2R=10k

vee

MOSFET_NMOSMOSFET6


INDQL6


BSIM3_Modelcmosn








RR1R=10k

VARVAR1

W=330L=0.6

EqnVar

INDQL1


vee

MOSFET_NMOSMOSFET1


CC2C=10 pF

INDQL7



V_DCSRC4Vdc=-2.5 V

Figure 18 Addition of a current mirror to the LNA to combine the differential outputs to a single output. Note care has to be taken as the additional current mirror will drop voltage and it may be necessary to raise the supply rail. Alternatively the cascaded LNA could feed a second stage differential amplifier (with current mir-ror to sum the outputs) where the differential signal could be combined.

Sheet

14 of 14

7 SUMMARY This tutorial gave the design equations to design a dif-ferential LNA using shunt feedback realized with on-chip inductors. A design example of a ‘bluetooth’ LNA was given, with the associated step-by-step design proc-ess to meet a given specification. ADS simulations have been given to predict the various circuit parameters of gain, noise figure and power consumption, all summa-rized in Table 3. Further discussions, described how the power gain could be further increased by adding another C-S stage to each ‘cascode’ stage output. This increased the power gain to >20dB and would greatly improve the noise figure of the receiver, as the noise figure contribution of the following mixer stage would be reduced by a 1/gain_LNA. Parameter Specifica-

tion Prediction Units

Frequency 2.45 to 2.85 2.45 to 2.85 GHz Noise Figure <3.5 2.8 dB Voltage Gain >20 27 dB Power Gain >10-15 11.4* dB Power con-sumption

<200 165** mW

Source/load impedance

50 50 ohms

Load Capaci-tance

0.4 0.4 pF

* Power gain > 20dB with another C-S stage added. ** Including ‘tail’ current mirror network not shown in the simulation schematics.

Table 3 Summary of simulated device per-formance.

Finally the definition of linearity was given, to-gether with the associated equations for the calcula-tion of 1dB gain compression point and 3rd Order Intercept point. A worked example was given, together with ADS Simulations showing close agreement with the hand calculations and the ADS simulations.

8 REFERENCES [1] T.H Lee, “The Design of CMOS Radio Fre-quency Integrated Circuits”, Cambridge University Press, ISBN 0 521 63922 0, Chapter 2. [2] C.S Kim, M Park, C-H Kim, Y C Hyeon, H K Yu, K Lee, K S Nam, “A fully integrated 1.9GHz

CMOS Low-noise amplifier” in IEE Microwave and guided wave letters, Vol 8, No 8 August 1998. [3] T Soorapanth, T.H Lee, “RF Linearity of Short-Channel MOSFETs”, IEEE Journal of Solid State Circuits, vol. 32, no. 5, May 1997

MOS Diff LNA - courses.e-ce.uth.gr · Sheet 2 of 14 Rs = 50Ω Lg M1 Using Agilent CMOS14 process...

Documents

Transcript of MOS Diff LNA - courses.e-ce.uth.gr · Sheet 2 of 14 Rs = 50Ω Lg M1 Using Agilent CMOS14 process...