You are on page 1of 131

RF Mixer Design

Modulation
Why?
Wired: superior shielding of coaxial lines at higher frequencies Wireless: antenna size FCC regulation: communication must occur in certain part of spectrum Allow simpler detection

Modulate?
Converts base-band signal to pass-band counterpart Varies certain parameters of carrier
07-21-2000 SiRACS-sglee 2

Detection
Demodulator
Inefficient under nNoise, nInterference, and nSmall input signal Require in advance nSignal amplification D RF amplifier nSelectivity D band-pass filter

07-21-2000

SiRACS-sglee

Detection
Antenna BPF RF amp Demodulator Baseband Output

RF amp requirements
High gain nAmplify signal for reliable processing Low noise nMinimize additional noise contribution High linearity nPrevent distortion or cross modulation Tuned nSelectively receive signals to reduce interference
07-21-2000 SiRACS-sglee 4

Receiver Design
RF amp problems
Oscillation n High gain RF amps tend to oscillate Must be tunable n Most receivers are not a fixed-frequency units Must design demodulator over frequency bands

07-21-2000

SiRACS-sglee

Detection
Filter

Task nFiltering narrow-band channel centered at high frequency nAlong with large interferers Problem nRequire prohibitively high Q - To provide 60 dB attenuation at 45 kHz nYet low loss filter
07-21-2000 SiRACS-sglee 6

Detection
Filter
Trade-off nOut-of-band rejection and nIn-band loss In-band loss being the more critical parameter Practical front-end BPF nCan only select the band of interest nPostpone channel selection to some other point

07-21-2000

SiRACS-sglee

Detection
Solution?
Super-heterodyne Concept nTranslate signal band to fixed lower frequencies nProgressively select and amplify - Allow relaxation in the Q of channel-select filter 9Filter Q = fc/f 9For a fixed Q, lower fc lower f nSelect signal with LO nDistribute gains nOperate demodulator at fixed frequency

07-21-2000

SiRACS-sglee

Heterodyne Receiver

LNA
Amplify RF signal & reduce mixer noise contribution

Image Reject Filter


Reject undesired out-of-band signals

Mixer
Allow channel selection and gain control at lower frequencies Allow fixed band IF filter nTune the LO frequency instead Dominates system performance nDetermines spec for LNA, LO, RF filter, IR filter, and IF stages
07-21-2000 SiRACS-sglee 9

Heterodyne Transceiver

07-21-2000

SiRACS-sglee

10

Mixer
Mixer
Misnomer, no mix rather multiply Perform frequency conversion Need nonlinear circuit component

Multipliers, modulators, and mixers, Multiplies signals


Multiplier nGenerates linear product of two signal w = k x y nHigh noise & low bandwidth Modulator nGenerate one of input multiplied by the sign of the other as an output w = y sign(x) Mixer nModulator, optimized for translation nOptimized for minimal noise and maximum linearity
07-21-2000 SiRACS-sglee 11

Mixer Specifications
Noise figure Linearity Conversion gain LO leakage LO input power Port matching Port isolation Power consumption

07-21-2000

SiRACS-sglee

12

Conversion Gain
Downconversion Mixer
Should provide sufficient power gain to compensate for IF filter loss and noise contribution from IF stage Too much gain may saturate mixer output (low IIP3) Gain specified in power nNF is power quantity nEasier to translate NF to IF stage to system NF

Power Gain
VO RS I O RS G= V R = I I L I RL
2 2

VO & VI: input and output voltages IO & II: input and output currents RL & RS: load and source resistance Increase RL two times 6 dB increase in voltage gain, but 3 dB increase in power gain
07-21-2000 SiRACS-sglee 13

Port Return Loss


RF and IF port matching
Necessary to avoid signal reflection and Excessive passband ripple in IF filter Typically, require more than 10 dB return loss

RF and LO port
Typically matched to 50 ohm LO port return loss can be relaxed Excessive reflection nSelf-mixing nSource of DC offset

IF port
Matched to that of IF filter

07-21-2000 SiRACS-sglee 14

Port Isolation
LO to RF feedthrough
LO signal leaking through antenna Should be small enough to avoid corrupting other RF systems

LNA & Mixer on same package


LO can feedthrough to RF input of LNA by passing RF filter and LNA Poor isolation between LNA out and mixer input makes IR filter useless

LO-to-IF & RF-to-IF isolation


Not important in Rx mixer High Q IF filter rejects it But large LO and RF signals at IF can saturate IF output port leading to poor P1dB

07-21-2000

SiRACS-sglee

15

Mixer Topologies
Mixer topologies
Conversion Gain nActive - Have conversion gain - Reduce IF stage noise contribution nPassive - Diode mixer, passive FET mixer, etc. - Conversion loss, high linearity, high LO power Device nDiode, FET, BJT mixer Frequency Conversion nUp- or down-conversion mixer LO drive nHigh- or low-side drive
07-21-2000 SiRACS-sglee 16

Mixer Topologies
Device nDouble-balanced: RF and LO driven differentially - Ring mixer, Gilbert cell mixer nSingle-balanced: RF or LO driven differentially - Dual-long tailed pair mixer nUnbalanced: RF and LO driven single-ended - Dual-gate mixer, exponential mixer, quadratic mixer LO-injection port (unbalanced) nGate mixer nSource mixer nDrain mixer
07-21-2000 SiRACS-sglee 17

Ideal Mixer
Fully-Balanced Switching Mixer
Vy = sin RF t
+1 Vy (RF) -1 Sx (LO) SxVy (IF)

S LO =

4 1 1 t t t sin sin 3 sin 5 / + LO LO LO 3 5 2 S IF = {cos( RF + LO )t + cos( RF LO )t 1 1 cos( RF + 3 LO )t cos( RF 3 LO )t 3 3 1 1 + cos( RF + 5 LO )t + cos( RF 5 LO )t/ } 5 5

Up-converter : choose RF + LO as IF Down-converter : choose RF - LO as IF RF > LO : low-side converter RF < LO : high-side converter
07-21-2000 SiRACS-sglee 18

Ideal Mixer
Note
Higher order frequency components No LO and RF at the IF

Image response
Even with ideal mixer, receiver suffers from image response Example n fLO = 1 GHz down converts nfRF = 0.9 GHz and fRFI = 1.1 GHz (image frequency) Image degrades the noise figure of the receiver Solutions nImage reject filter ahead of mixer nImage rejection mixer nDirect conversion receiver
07-21-2000 SiRACS-sglee 19

Unbalanced Mixer
Simplest
Possible to achieve lowest noise figure No LO and RF isolation at IF need capacitor at IF to bypass them

Low noise implementation


Need LO and RF signal to have small IF noise power Degenerate driver stage with capacitor

07-21-2000

SiRACS-sglee

20

Square-Law Mixer
Gate mixer
vRF vLO vIF

iD =

(Vgs VT ) 2 2

Vgs = vRF cos( RF t ) + vLO cos( LO t ) + VGS


(Vgs VT )2 = (vRF cos( RFt) + vLO cos( LOt) +VGS VT )2 = [vRF cos( RFt) + vLO cos( LOt)]2 + (VGS VT )2
2 2 vRF v + vLO v = + (VGS VT )2 + RF cos( 2RFt) + LO cos( 2LOt) 2 2 2 DC 2-nd harmonic + vRF vLO cos{( RF LO)t}+ vRF vLO cos{( RF +LO)t}

Mixing terms

Gc =

vLO 2
SiRACS-sglee

CoxW 2L
21

07-21-2000

Single-Balanced Mixer
Compare to DBM
Simple, half the current, more head room Less dependent on LO balance Easy RF port matching Fewer noise contributors

Inductive degeneration
Z in = rb + Z e + 1 sC + T Z e s
Good noise performance Boosts class AB behavior

Common-base driver stage


Tend to be more linear Low power gain (due to no current gain) Higher NF : more noise contribution bySiRACS-sglee switching tr 22

07-21-2000

Single-Balanced Mixer
iO1 iC2 Q2 iC3 Q3 iO2

For ideal switching


iC1 = I C1 + g m vRF cos RF t SC 2 = 1 2 + cos LO t + O(3 LO ) 2 1 2 SC 3 = cos LO t + O(3 LO ) 2 I g v 2I 2g v iO1 = iC1 SC 2 = C1 + m RF cos RF t + C1 cos LO t + m RF cos RF t cos LO t + / 2 2 I g v 2I = C1 + m RF cos RF t + C1 cos LO t 2 2 g v g v + m RF cos( LO RF ) + m RF cos( LO + RF ) + / I g v 2I 2g v iO 2 = iC1 SC 3 = C1 + m RF cos RF t C1 cos LO t m RF cos RF t cos LO t + / 2 2 I g v 2I = C1 + m RF cos RF t C1 cos LO t 2 2 g v g v m RF cos( LO RF ) m RF cos( LO + RF ) + / iO = iO1 iO 2 = 4 I C1 2 2 cos LO t + g m vRF cos( LO RF ) + g m vRF cos( LO + RF ) + /
SiRACS-sglee 23

VLOcosLOt

VRFcosRFt iC2

iC1 Q1

t iC3 t iC1

IC1 t

07-21-2000

Single-Balanced Mixer
Inherent loss
2/ (~ -3.9 dB) by the frequency conversion RF power converted to IF, RF+IF, and higher order --

No LO & RF isolation at IF
Limits the usage for down-conversion only nIF filter attenuates LO & RF Can saturate IF output can degrade P1dB Parallel capacitors at the IF helps

07-21-2000

SiRACS-sglee

24

Single-Balanced Mixer
Single-ended output
Noise figure degradation nUp to 6.9 dB by the direct feedthrough of driver IF noise nDC component in LO No isolation between RF and IF port at IF frequency IF Inductive degeneration IF noise amplified with higher gain

Differential output
Add IF signal Cancel RF signal (common-mode signal) Add LO signal (diff-mode signal) Cancel IF noise feedthrough from driver stage nSignificant NF reduction nMust be differential

07-21-2000

SiRACS-sglee

25

Double-Balanced Mixer
Q1 & Q2
Diff pair driver stage Amplifies RF signal Reduce noise contribution from switching quad Should be large for small rb: low noise

Q1, Q2, Q3, & Q4 IT


Diff switching quad Should be optimized for noise of Q1 & Q2 Input-referred noise by IC decrease with IT While by IB increase with IT

Ze
Improve linearity, reduce gain L is best for linearity and NF Higher gain minimize quad noise contribution
SiRACS-sglee 26

07-21-2000

Double-Balanced Mixer
iO1 iC3 Q3 vLOcosLOt iC1 Q1 iC2 Q2 iC4 Q4 iC5 Q5 iC6 Q6 iO2

vRFcosRFt

1 1

iC3 = iC6 t iC4 = iC5 t iC1

IC1 iC2 IC2


07-21-2000

t t

g m vRF cos RF t 2 g v g v iC 2 = I C + m RF cos( RF + )t + / = I C m RF cos RF t + / 2 2 1 2 SC 3 = SC 6 = + cos LO t + / 2 1 2 1 2 SC 4 = SC 5 = + cos( LO + )t + / = cos LO t + / 2 2 1 iO1 = iC1S C 3 + iC 2 S C 5 = I C + 2 g m vRF cos RF t cos LO t + / 1 1 = I C + g m vRF cos( LO RF )t + g m vRF cos( LO + RF )t + / 1 iO 2 = iC1S C 4 + iC 2 SC 5 = I C 2 g m vRF cos s t cos ot + / 1 1 = I C g m vRF cos( LO RF )t g m vRF cos( LO + RF )t + / 1 1 iO1 = I C + g m vRF cos( LO RF )t + g m vRF cos( LO + RF )t + / 1 1 iO 2 = I C g m vRF cos( LO RF )t g m vRF cos( LO + RF )t + / 2 2 iO1 iO 2 = g m vRF cos( LO RF )t + g m vRF cos( LO + RF )t + / iC1 = I C +
SiRACS-sglee 27

Double-Balanced Mixer
RF & LO isolation at IF Require balun for differential LO and RF Same conversion for twice the ICC of SBM

Differential output
Increase output power Cancel DC

Larger LO
Closer to ideal switching nIncrease conversion gain nReduce noise figure Excessive current pumped into quad nGenerate additional IM3 nReduce voltage headroom nIncrease power consumption nBipolar: VLO ~ 300 mV = 0 dBm is used for ideal switching
07-21-2000 SiRACS-sglee 28

Double-Balanced Mixer
Linearity
High even-order linearity

Differential LO drive
Transformer nPower loss nExternal component nBulky Active balun nExtra circuitry nAdditional noise: IF noise pumped into quad show up at IF - Capacitive degeneration helps nRequire high power balun for efficient switching Prefer single-ended LO

07-21-2000

SiRACS-sglee

29

Double-Balanced Mixer
Single-ended LO
equivalent representation

VS

VS/2

com. mode diff. mode


RS

VS/2 VS/2

VS/2 RS VS/2 + VS/2 = VS

VS/2 - VS/2 = 0

LO leakage nSerious with up-mixer, must be removed nLeads to additional output buffer Solutions? nCan be related to common-mode gain of switching quad nCascode helps

07-21-2000

SiRACS-sglee

30

Double-Balanced Mixer
2LO harmonics at output
Often serious problem nLarge amplitude saturates switching quad nDetermine by the ratio of load impedance and driver stage output impedance nCascode helps

Input impedance
Z in = 2(rb + Z e + 1 sC + T Z e s )

nHigher input impedance than SBM nCapacitive Ze generates negative resistance


07-21-2000 SiRACS-sglee 31

MOSFET Double-Balanced Mixer


Overall
Similar to BJT version Hard to achieve conversion gain

Linearity
Better linearity (?) Higher (VGS - VT) helps driver stage linearity

Switching quad
Need larger LO to switch transistor Need larger LO to minimize noise contribution

07-21-2000

SiRACS-sglee

32

Switching Efficiency of CMOS DBM


Iod

Incomplete switching
ISS

2Vp

Q3 LO

Q4

Q5

Q6
2ISS/ 2ISS/

ISS/2 Vid

Q1 RF

Q2

-ISS

Transfer characteristic of a CMOS differential amplifier


2

(Square IV) 1 (Linear IV)

0 0.1 1 10

aLO/2VP

Switching efficiency dependence on LO amplitude from numerical calculation


07-21-2000

- Quadratic I-V: 1dB loss point compared to large-LO case is at ALO = 2Vp, and 3dB loss point at ALO = 1.4Vp - Linear I-V: 1dB loss point compared to maximum value is at ALO = 4.4Vp,, and 3dB point at ALO = 2.7Vp - For BJT balanced mixer, 1dB loss point compared to large-LO case is at ALO = 100 mV
SiRACS-sglee 33

x 1/

Noise Figure
SSB noise figure
Applicable to heterodyne receiver Information only on one of sideband

DSB noise figure


Applicable to homodyne receiver Both sideband of LO are converted to baseband

Heterodyne Twice as many noise contributor


07-21-2000 SiRACS-sglee

Homodyne

34

Noise Figure
Noise Figure Meter Noise Figure Meter

Noise Source LO

Noise Source

BPF1

BPF2

LO

BPF2

DSB Measurement
NFDSB NFSSB

SSB Measurement
(1) ( 2)

N a , DSB Si N i Si N i = = = 1+ S o N o 2 S i Ga (2 N i Ga + N a , DSB ) 2Ga N i N a , DSB Si N i Si N i = = = 21 + S o N o S i Ga (2 N i Ga + N a , DSB ) 2Ga N i

Noise less mixer: NFSSB = 3dB and NFDSB = 0dB


07-21-2000 SiRACS-sglee 35

Noise Figure
Noise Figure Meter Noise Figure Meter

Noise Source LO

Noise Source

BPF1

BPF2

LO

BPF2

DSB Measurement
NFDSB = NFSSB =

SSB Measurement
(1)

N a , DSB N a , DSB N Si N i Si N i = = 1+ a , DSB S o N o Si (Ga + Gi ) (N i (Ga + Gi ) + N a , DSB ) (Ga + Gi ) N i (Ga + Gi ) N i 2Ga N i N N Si N i Si N i = = 1 + a , DSB a , DSB S o N o S i Ga (N i Ga + N a , DSB ) Ga N i Ga N i ( 2)

The SSB NF is ~ 3dB higher than DSB NF if signal and image


bands experience equal gain The SSB and DSB NF is same for Ga >> Gi; the narrow band mixer
07-21-2000 SiRACS-sglee 36

SSB NF Measurement
Noise Figure Meter Noise Source BPF1

LO

BPF2

Calibrate NFM at IF frequency BPF1: need for single-side band meas. BPF2: stop the noise contribution from the LO signal generator Minimize the cable & BPF1 loss, or subtract the loss from the measured NF Check the ENR number Watch for the linear IF amp configuration - need BPF1 Watch for the RF matching effect on the noise figure test with line-stretcher
07-21-2000 SiRACS-sglee 37

Mixer Noise
Suffer from high noise figure
Higher than amplifiers Noise transferred from multiple frequency bands to output

Periodically time-varying statistics


Operating point of the devices changes periodically Transfer function for signal and noise change periodically with time Periodic random process is called cyclostationary

Noise figure simulation


Require nonlinear noise simulator Commercial models with high accuracy is not yet available Analytic approach with simplifying assumptions: Terrovits et al., JSCC 1999 / Darabi et al., IEEE SSC 2000

07-21-2000

SiRACS-sglee

38

Noise in SBM
Noise contributors
Driver stage noise, M3 Switching pair thermal noise, M1 & M2 LO signal noise Thermal noise from load resistors 1/f noise: significant in direct conversion mixer

07-21-2000

SiRACS-sglee

39

Noise in SBM
1) Driver stage noise from M3
Cyclostationary due to the switching action of quads

81%

9%

10%

Noise contribution in the ideal switching


07-21-2000 SiRACS-sglee 40

Noise in SBM
2) Switching pair thermal noise from M1 & M2
White and cyclostationary Noise contribution from near even LO harmonics Large LO, noise contribution of switching pair is lower than driver stage

07-21-2000

SiRACS-sglee

41

Noise in SBM
3) LO signal noise
Depends on oscillator noise, cyclostationary Bandwidth

4) Thermal noise from load resistors


White and bias independent

5) 1/f noise
Significant in FETs

07-21-2000

SiRACS-sglee

42

Noise in SBM
NFSSB

( 3 + rg 3 g m 3 ) g m 3 + 2 1 G + ( RLO + 2rg1 ) G + 1 / RL = 2+ 3 c c2 gm 3 Rs 2{( 3 + rg 3 g m3 ) g m3 + 2 1 G + ( RLO + 2rg1 ) G } + 1 / RL + 2 3 c c2 gm 3 Rs


__ ___ 2

__

___ 2

SBM

NFSSB =

DBM

c: conversion gain efficiency, maximum 2/ : driver stage noise transfer efficiency, maximum 1 i: drain channel noise coef for i-th FET rgi: gate resistance of i-th FET G & G2: time and squared time average of G(t) = 2gm1gm2/(gm1+gm2) Monotonic increasing function of current and LO amplitude Capacitor effects are neglected

07-21-2000

SiRACS-sglee

43

Noise in SBM
Lower noise for ideal switching Strong correlation with conversion gain For same GC , DBM dissipates twice the power of SBM, higher NF

07-21-2000

SiRACS-sglee

44

Image Response
Shift by 90o in time domain: x x-T/4 Shift by 90o in frequency domain: Hilbert transformation
+ j H ( ) = j sign( ) = j for < 0 for > 0

Expressions of sine and cosine

sin(t )

H ( ) = cos(t )

1 j ( e + e + j ) 2 j sin( ) = (e j e + j ) 2 cos( ) =
Shift by 90o in (a) time and (b) frequency domain

07-21-2000

SiRACS-sglee

45

Image Reject Mixer


Hartley architecture

ARF cos( RF t ) + Aim cos( imt )

- Image rejection improves with larger LO - Image rejection depends more strongly on phase mismatch
07-21-2000 SiRACS-sglee 46

Image Reject Mixer


Weaver architecture

XA()

- Image rejection improves with larger LO - Image rejection depends more strongly on phase mismatch
SiRACS-sglee 47

XB( )

07-21-2000

Image Reject Mixer


Newer architecture
No LPF Mismatches self-compensates

07-21-2000

SiRACS-sglee

48

Balun/180-Degree Power Combiner


Out In Out In Out Out

out

out

In

0o

out

in

out

out
in
in

0o

180o

180o

0o

out

In

in

180o

L =
07-21-2000 SiRACS-sglee

1 = 2R C
49

Balun/180-Degree Power Combiner


RF+ R11 RFR12 Q13
Out In

RF_out Q14

In

L i1-i2 C

out

In

In

i1

i2

=
07-21-2000 SiRACS-sglee

1 C 2L
50

Mixer Linearization Technique


Two-Tone 3rd-Order Intercept Point (IP3)
For memoryless system, in/out transfer characteristics is given as
vo (t ) = 1vi (t ) + 2 vi (t ) 2 + 3vi (t ) 3 + O[vi (t ) 4 ]
with
1 LO , 2 LO = RF LO , RF LO + = IF , IF +

vi = vRF sin(1t ) + vRF sin( 2t ) + vLO sin( LO t ) = vRF sin( RF t ) + vRF sin([ RF + ]t ) + vLO sin( LO t )

21 2 LO ,2 2 1 LO = RF LO , RF LO + 2 = IF , IF + 2

4 1 IIP3(dBV ) = 10 LOG10 3 3

4 13 OIP3(dBV ) = 10 LOG10 3 3

Therefore linearization circuit should provide both smaller |3| and no reduction in voltage gain (1).
07-21-2000 SiRACS-sglee 51

Mixer Linearization Technique


Nonlinearity of an FET
5 4 25 20 15 10 5 0 -5 -10 -15 0 1 2 3 3 2 1 0 -1 -2

ids

vi

V-I Characteristics

Ids gm gm' gm"

ids = g m vi +

g g m 2 vi + m vi3 2! 3!

-3

Gate-to-Source Voltage [V]

- High Vgs D high gm and small |gm| D Linear Amplification - Low Vgs D low gm and large |gm| D Distortion generator
07-21-2000 SiRACS-sglee 52

gm' [mA/V2] and gm" [mA/V3]

Ids [mA] and gm [mA/V]

Mixer Linearization Technique


Resistive emitter degeneration

' gm = g m /(1 + )

, where = g m Re

Incremental gain vs. Vin

- Smaller gain and higher order coefficients for larger - Identify the optimum value for the maximum linearity - Not effective for CMOS mixer due to insufficient gm - Inductive/capacitive degeneration for low voltage operation
07-21-2000 SiRACS-sglee 53

Mixer Linearization Technique


Inductive emitter degeneration : Fong et al.

- No additional noise by inductor - Class AB operation - More effective for P1dB than IP3 (IP1dB=-1.5 dBm, IIP3=+2.5dBm) - Good for CMOS technology?
07-21-2000 SiRACS-sglee 54

Mixer Linearization Technique


Multi-tanh Mixer

Incremental gain vs. Vin

- Smaller gain for A not equal to 1 - Identify the optimum A value for the highest linearity - Not effective with CMOS mixer due to insufficient gm (?)
Incremental gain vs. Vin
07-21-2000 SiRACS-sglee 55

Mixer Linearization Technique


MicroMixer

(a) Incremental gain, (b) RIN, (c) current of Q1 & Q3, (d) input-referred noise

- Q1 can handle unlimited amounts of current during negative vin - Same with Q2, Q3 during positive excursions of vin - No gain compression: rather, the gain increases at high input levels class AB - Good for CMOS
07-21-2000 SiRACS-sglee 56

Mixer Linearization Technique


Resistive Mixer

- For FET operated at linear region, channel is a linear voltagecontrolled resistor - Potentiometric mixer, balanced mixer, etc. - Highly linear, but exhibits inherent loss: in case of additional gain block, linearity degraded due to IF or RF gain block
07-21-2000 SiRACS-sglee 57

Low-Voltage-Operation Mixer

- Parallel LC resonance at RF frequency - Area occupied by the inductor - Raytheon patent


07-21-2000

- Fraello et al., IEE EL , 1997

SiRACS-sglee

58

Low-Voltage-Operation Mixer

- G. Giustolisi , IEEE , 1998

07-21-2000

SiRACS-sglee

59

Low-Voltage-Operation Mixer

real[ Z in ] ~

gm 2C gs CEE

- Razavi, ISSCC 1996 - Capacitive source degeneration for low voltage operation - Stability : possible negative input resistance at RF input ports - Relatively high noise figure(?)
07-21-2000 SiRACS-sglee 60

High Conversion-Gain Mixer

- Schmatz et al., GaAs IC Symp. - Current splitting with common-mode feedback

07-21-2000

SiRACS-sglee

61

High Conversion-Gain Mixer

- MacEachern et al., IEEE, 1998 - LC resonance at RF frequency - Charge-injection method with common-mode feedback

07-21-2000

SiRACS-sglee

62

CMOS Mixer
- Jacques C. Rudell , IEEE J. Solid-State Circuits, Dec. 1997

07-21-2000

SiRACS-sglee

63

CMOS Mixers
Arvin R. Shahani et al., IEEE J. Solid-State Circuits, Dec. 1997 Double-balanced resistive mixer

07-21-2000

SiRACS-sglee

64

CMOS Mixers
- Andrew N. Karanicolas, IEEE J. Solid-State Circuits, Dec. 1996

07-21-2000

SiRACS-sglee

65

CMOS Mixers
- Seungwook Lee , IEEE, 1997
- Ahmadreza Rofougaran , IEEE J. Solid-State Circuits, July 1996

07-21-2000

SiRACS-sglee

66

CMOS Mixers
- Litmanen et al., IEEE, 1998

07-21-2000

SiRACS-sglee

67

MOS SB Down Converter


H. Wang, IEEE ISSCC, Feb. 1998

07-21-2000

SiRACS-sglee

68

MOSFET Mixer
Current-reused bleeding mixer
VDD

VDD

R1 IF+

IBIAS2 IF-

R2

R1
+ I OUT

R2 IF+ IFM2 M4
I OUT ,bld

,bld

LO+

M5

M2

M6

M7

M4

M8

LO+

M3

LO-

LO-

Cbypass
RF+ M1 M3 RF-

RF

M1

IBIAS1

07-21-2000

SiRACS-sglee

69

Mixer Fundamentals & Active Mixer Design


 

References
x

Keng Leong Fong and R. G. Meyer, Monolothic RF Active Mixer Design, IEEE Trans. Circuits and Systems - II: Analog and Digital Signal Processing, vol. 46, no. 3, March 1999 Mixer Fundamentals and Active Mixer Design, B. Gilbert, EPEL Electronics Laboratories Advanced Engineering Course in RF Design for Wireless Communications Systems, Lausanne, 1996 Communication Circuits: Analysis and Design, K. K. Clarke and D. T. Hess, AddisonWesley, 1971

04-09-98

SGL/EE/ICU

Modulation and Detection


s

Reasons for modulation


x x x x

Wired: superior shielding of coaxial lines at higher frequencies Wireless: antenna size FCC regulation: communication must occur in certain part of spectrum Allow simpler detection

Signal
x x

baseband: spectrum near = 0 passband: spectrum near carrier freq. c

Modulation
x x

Converts baseband signal to passband counterpart Varies certain parameters of carrier

04-09-98

SGL/EE/ICU

Modulation and Detection


s

Important aspect of modem


x

Quality (e.g., SNR)

Quality depends on the type of modem for a given channel attenuation and transmitted power Limited bandwidth for wireless application Modulation affects required bandwidth Linearity requirement determines the kind of power amplifier Typically, < 

Spectral efficiency

Power efficiency

Modem
x

Analog

Process of converting the baseband signal to passband and vice versa Signal processing: channel coding, decoding, spreading, despreading, etc. Represented by similar numerical expression as in analog modem

Baseband

04-09-98

SGL/EE/ICU

Wireless Communication Environment


s

Hostile Limited spectrum allocated


Limited rate of information coding, compression, bandwidth efficient modulation
x

RF section
x

Transmitter

Employ narrow-band modulation Amplification Filtering to avoid leakage to adjacent channels Antenna matching Desired channel selection Undesired signal rejection Amplification Demodulation Error detection and/or correction Information conditioning and output
SGL/EE/ICU 5

Receiver

04-09-98

General Considerations
s

Interference rejection
x

Require extremely high-Q filter to provide 60 dB attenuation at 45 kHz


With simple LC filter require Q ~ 107 Prohibitive Filter  For a fixed Q L&C,
higher Q filter, higher-order filter, more L&C, more loss


-order filter cascade of  1st-order filter Trade-off between loss and Q

Typical filters

04-09-98

SGL/EE/ICU

General Considerations
x

Choice of BPF filter


Trade-off between out-of-band rejection and in-band loss In-band loss being the more critical parameter Band: entire spectrum of a particular standard Channel: signal bandwidth of only one user

Practical front-end BPF


Can only select the band of interest Postpone channel selection to some other point

04-09-98

SGL/EE/ICU

Receivers
Receiver design
Antenna RF amp Demod BB Output

Demodulator
x x x

Inefficient under noise, interference, and small input signal Require signal amplification and selectivity in advance Introduced RF amplifier

RF amplifier Requirements
x x x x

High gain: amplify signal for reliable processing Low noise: minimize additional noise contribution High linearity: prevent distortion or cross modulation Need to be tuned: selectively receive signals to reduce interference

04-09-98

SGL/EE/ICU

Receivers
s

Problems
x x

High gain RF amp tends to oscillate Most receivers are not a fixed-frequency units need to tune the high-gain RF amp over wide frequency range Need to design demodulator to have adequate performance over frequency bands

Solution? Superheterodyne Concept


x

x x x

Translate incoming signal to new, fixed intermediate frequency Select signal with LO Distribute gains Operate demod at fixed freq

Antenna

RF amp

Mixer

LO

IF amp

Demod

BB Output

04-09-98

SGL/EE/ICU

Heterodyne Receivers
s

Task
x x

Filtering narrow band channel centered at high frequency Along with large interferers Require prohibitively high Q yet low loss filter Translate signal band to lower frequencies And progressively select and amplify Allow relaxation in the  of channel-select filter Filter  For a fixed , lower lower  Mixer: downconversion LPF: remove upconverted signal LNA: precede due to the mixers high NF 0: generated by local oscillator 2: called intermediate frequency (IF)

Problem
x

Solution
x x x

x x x x x

04-09-98

SGL/EE/ICU

10

Heterodyne Receivers
s

Mixer
x

Allow channel selection and gain control at lower frequencies

High Q filters and AGC can be implemented economically Tune the LO frequency instead Its performance dominates system performance Affects performance requirement of LNA, LO, RF filter, IR filter, and IF stages

Allow fixed band IF filter

Very important building block


LNA
x

Amplify RF signal to reduce noise contribution from mixer Reject undesired out-of-band signals

Image Reject Filter


x

04-09-98

SGL/EE/ICU

11

Mixer
s

Mixer
x x x x

Misnomer, no mix rather multiply Circuit element that performs frequency conversion Nonlinear circuit component Tends to dominate receiver performance

Multipliers, Modulators, and Mixers


x x

All of them multiplies signals Multiplier

Generates linear product of two signal  = k  High noise & low bandwidth Generate one of input multiplied by the sign of the other as an output  = sign() Modulator, optimized for translation Optimized for minimal noise and maximum linearity

Modulator

Mixer

04-09-98

SGL/EE/ICU

12

Mixer
s

Mixer topologies
x

Conversion Gain

Active
Have conversion gain Reduce IF stage noise contribution

Passive
Diode mixer, passive FET mixer, etc. Conversion loss, high linearity, high LO power

Topology and Isolation

Double-balanced: RF and LO driven differentially


Ring mixer, Gilbert cell mixer

Single-balanced: RF or LO driven differentially


Dual-long tailed pair mixer

Unbalanced: RF and LO driven single-ended


Dual-gate mixer, exponential mixer, quadratic mixer

Device

Schottky diode, FET, BJT, or CMOS mixer Up- or down-conversion mixer High- or low-side drive
SGL/EE/ICU 13

Frequency Conversion

LO drive

04-09-98

Ideal Mixer Theory


s

Fully-Balanced Switching Mixer


=  
+1  -1  
     =    
 
+  
!     = {  + 
+   
    +  
   
    +   +  
+    
!}  

x x x x

Up-converter : choose  +  as  Down-converter : choose  -  as   : low-side converter  : high-side converter

04-09-98

SGL/EE/ICU

14

Ideal Mixer Theory


s

Note
x x

Higher order frequency components No  and  at the  Even with ideal mixer, receiver suffers from image response Example

Image response
x x

Assume  = 1 GHz and   = 0.9 GHz  =  -   = 1 GHz - 0.9 GHz = 0.1 GHz Another IF  =   -  = 1.1 GHz - 1 GHz = 0.1 GHz   : image frequency

x x

Image response degrades the noise figure of the receiver Solutions


Include image reject filter ahead of mixer Design image rejection mixer : requires two mixer cells operating in quadrature

04-09-98

SGL/EE/ICU

15

Receiver Performance Parameters


s

Sensitivity
x x

Smallest received signal receiver needs to achieve a specific bit error rate (BER) Depends on system noise figure and demodulation scheme Ability to detect desired signal and reject undesired signal Includes

Selectiviey
x x

Adjacent channel selectivity,


Depends on IP3 of LNA and Mixer, selectivities of IF and BB filters, LO phase noise

Image rejection,
Depends on selectivities of RF and IR filters

Out-of-band blocker rejection,


Depends on selectivities of RF and IR filters, desensitization effect on LNA and down-conversion mixer by blocker, LO phase noise

Overloading
x x

Largest desired signal receiver can handle while maintaining a specific BER Depends on system P1dB
SGL/EE/ICU 16

Power Consumption: determine the usage time of portable receiver

04-09-98

Noise Figure

 =

        + + +                = +    

x x

Need sufficient  to reduce  Low noise mixer is highly desirable to relax 

04-09-98

SGL/EE/ICU

17

Single- and Double-Sideband Noise Figure


s

SSB Noise Figure


x x

Applicable to heterodyne receiver Only one of the sideband is converted to IF (image is rejected) Applicable to homodyne (direct conversion) receiver Two sidebands of LO signal are converted to baseband

DSB Noise Figure


x x

Heterodyne
Twice as many noise contributor
04-09-98 SGL/EE/ICU

Homodyne
18

Single- and Double-Sideband Noise Figure


DSB Measurement

    

SSB Measurement

     

 = 

             = =  +                +            

              = = = +           +          

For  >> ,  is about 3 dB higher than 


04-09-98 SGL/EE/ICU 19

NF Measurement
s

Noise Figure(SSB)

     

x x x

Calibrate NFM at IF frequency BPF1: need for single-side band meas. BPF2: stop the noise contribution from the LO signal generator Minimize the cable & BPF1 loss, or subtract the loss from the measured NF Check the ENR number Watch for the linear IF amp configuration - need BPF1 Watch for the RF matching effect on the noise figure test with line-stretcher

x x x x

04-09-98

SGL/EE/ICU

20

Conversion Gain
s

Downconversion Mixer
x

x x

Should provide sufficient power gain to compensate for IF filter loss and noise contribution from IF stage Too much gain may saturate mixer output (low IIP3) Gain specified in power

NF is power quantity Easier to translate NF to IF stage to system NF

Power Gain
     =   =     
x x x x
 

 & : input and output voltages  & : input and output currents
&

: load and source resistance Increase


 two times 6 dB increase in voltage gain, but 3 dB increase in power gain

04-09-98

SGL/EE/ICU

21

Gain Compression

 >  
x x x x x

Desired signal distorted at the mixer output Cause AM to PM modulation No information lost if is frequency modulated If is phase modulated, unwanted phase shift due to AM-to-PM conversion Lead to detection error degradation (BER) Distortion of has no effect on system performance can overload mixer Cause gain compression on the desired signal increase IF noise contribution
SGL/EE/ICU 22


>  
x x x

04-09-98

3rd-Order Intermodulation Distortion

Low 
x

IM3 dominated by 3rd-order nonlinearity Higher order nonlinearities become more important Depends on phase relationship among odd-order nonlinearities, IM3 may increase at higher or lower rate than the cube of Coherent addition of IM3 from LNA and mixer:

High 
x x

    =  +        

IM3 products add incoherently(out-of-phase):

 = 

   +       

IP3: 9.6 dB higher than P1dB at low frequency, higher in many practical design except the class AB mixer
SGL/EE/ICU 23

04-09-98

Power Consumption
s

Optimize mixer power consumption to avoid increase in other building blocks


x x x x x x x

Mixer with high NF Increase gain requirement for LNA Increase power consumption of LNA Increase IP3 requirement for mixer Increase mixer power consumption Require higher LO drive Increase LO buffer bias current

04-09-98

SGL/EE/ICU

24

Port Return Loss


s

RF and IF port matching


x x x

Necessary to avoid signal reflection and Excessive passband ripple in IF filter Typically, require more than 10 dB return loss Typically matched to 50 ohm LO port return loss can be relaxed Excessive reflection

RF and LO port
x x x

Self-mixing Source of DC offset

IF port
x

Matched to that of IF filter

04-09-98

SGL/EE/ICU

25

Port Isolation
s

LO to RF feedthrough
x x

LO signal leaking through antenna Should be small enough to avoid corrupting other RF systems LO can feedthrough to RF input of LNA by passing RF filter and LNA Poor isolation between LNA out and mixer input makes IR filter useless Not important in Rx mixer High Q IF filter rejects it But large LO and RF signals at IF can saturate IF output port leading to poor P1dB

LNA & Mixer on same package


x x

LO-to-IF & RF-to-IF isolation


x x x

04-09-98

SGL/EE/ICU

26

Double-Balanced Mixer
s

 & : diff pair driver stage


x x

x
s s

Amplifies RF signal Reduce noise contribution from switching quad Should be large for small : low noise

, , , & : diff switching quad 


x x x x x

Improve linearity, reduce gain Can be implemented with , , or   & : lower NF L gives best linearity Higher gain minimize quad noise contribution Should be optimized for noise of  &  Input-referred noise by decrease with While by  increase with 


x x x

04-09-98

SGL/EE/ICU

27

Double-Balanced Mixer
     
          

   

   



 

             =   +   (  + ) + ! =        + !       =   = +    + !        =   = + (  + ) + ! =   + !     =   +    =   +          + !   =   +    (   ) +    (  +  ) + !    =   +    =            + !   =      (   )    (  +  ) + !  =   +
SGL/EE/ICU 28

04-09-98

Double-Balanced Mixer
  =  +  
x x x

  
  (   ) + 
  (  +  ) + !   =  
  (   ) 
 (  +  ) + !

Assumed ideal instantaneous switching RF & LO isolation at IF Require extra circuitry and supply current for balanced LO and RF drive for balanced operation Converted signal adds Increase output power Cancel the common-mode noise from Cancel DC Due to power lost in (RF + RF) term and other higher frequency terms

Differential output
x x x

1/ factor
x x

04-09-98

SGL/EE/ICU

29

Double-Balanced Mixer
  = ( +   +   +     )
x
s

Capacitive  generates negative real number Assume const gain and noise power at output of driver Instant switching lead to 3.9 dB increase in NF + noise contribution from switching pairs High side mixing gives better NF than the low side mixing Low NF Excessive current pumped into quad generate additional IM3 Decrease voltage head room Increased power consumption Bipolar:  ~ 300 mV = 0 dBm used for low NF Reasonably large for low Too large Cje too much current into quad linearity degradation
SGL/EE/ICU 30

Noise characteristics
x x x

Inductive degeneration
x

Large LO signal
x x x x x

Quad Tr. Size


x x

04-09-98

FET Double-Balanced Mixer


s

FETs
x

MESFET or MOSFET Similar to the case of BJT version Better linearity (?) Driver stage can be improved by higher ( 
 ) Needs larger LO to turn off transistor Need larger LO to minimize noise contribution

Overall
x

Linearity
x x

Switching quad
x x

04-09-98

SGL/EE/ICU

31

Single-Balanced Mixer
s

Compare to DBM
x x x x

Simple, half the current, more head room Performance less dependent on LO balance Easy RF port matching Fewer noise contributers Good noise performance Exhibit class AB behavior High P1dB with low bias current

Inductive degeneration
x x x

  =  +   +driver   stage +     Common-base


x x x

Tend to be more linear Low power gain (due to no current gain) Higher NF : more noise contribution from quad

04-09-98

SGL/EE/ICU

32

Single-Balanced Mixer
       
 =   +         +    + !        = + (  + ) + ! =    + !          =   =  +      +     +         + !       =  +      +           +   (   ) +   (  +  ) + !         =   =  +                  + !       =  +                  (   )   (  +  ) + !   =



 











04-09-98

SGL/EE/ICU

33

Single-Balanced Mixer
x
s

Assumed ideal instantaneous switching No problem, IF filter attenuates them Can saturate IF output can degrade P1dB Parallel capacitors at the IF Increase noise figure by 6.9 dB for constant output noise power of driver stage DC component in LO IF noise from driver stage directly show up at IF With inductive degeneration, IF noise amplified with high gain IF power adds Cancel RF but not the LO Significant NF reduction: has to be differential output for low NF Quad Tr. IF noise mix with RF signal DC Increase noise at the IF port With LO buffer, bandpass or highpass load help reduce the LO noise contribution

No LO & RF isolation at IF
x x x

Single-ended output
x x x

Differential output
x x x

Noise from LO
x x x

04-09-98

SGL/EE/ICU

34

DB Mixer Noise Sources


Noise source from RF transistor
x x

 () : have no control  () : have no control, unless insert image reject filter between transconductance stage and mixer core  () : depends on the way signal applies;

Differential mode: becomes feedthough and cancels out at each IF Common-mode: shows up at IF, cancels when outputs taken differentially

Noise source from current source


x x

 () : feedthrough to IF, cancel out if outputs taken differentially  () and  () : feed to mixer core as common-mode signal, down converts and show up at IF. Can be cancelled out if outputs taken differentially Same as SBM

Noise source from LO transistors


x

04-09-98

SGL/EE/ICU

35

SB Mixer Noise Sources


Noise source from RF transistor
x x

 () : unavoidable, determined by device characteristic  () : image noise component, possible to stop using band-pass filter or resonance (trap) circuit  () : show up at the IF due to the lack of RF-IF isolation. Possible to cancel by differential output or stop before mixer core Problem due to non-ideal switching mixer core becomes differential amplifier multiplier mathematics complicated Solution?: faster switch, higher LO drive LO buffer can also add extra noise Transformer: wide band, bulky, expensive, lossy LC power combining circuit: narrow band, require many passive components Active circuit realization: biasing, extra noise, matching

Noise source from LO transistors


x

x x
s

Differential power combining approach


x x x

04-09-98

SGL/EE/ICU

36

Mixer Linearity Improvement Techniques


s

To Improve linearity
x x x

Higher current and large headroom Distribute voltage to avoid none of the transistors fall to early saturation (or cutoff) Choose the right circuit topology

04-09-98

SGL/EE/ICU

37

Unbalanced Mixer
s

Simplest
x x

Possible to achieve lowest noise figure No LO and RF isolation at IF need capacitor at IF to bypass them Need LO and RF signal to have small IF noise power Degenerate driver stage with capacitor

Low noise implementation


x

04-09-98

SGL/EE/ICU

38

Characterization
s

Matching
x x

Measure s-parameters using VNA with other ports 50 terminated Watch for linear vs. nonlinear s-parameters Typically use spectrum analyzers Voltage vs. power conversion gain

Conversion Gain
x x

RS = RL, same If not, varies (need to calculate)

IP3 and P1dB


x x x

Typically use spectrum analyzers Require 3 signal generators for IP3 Measure IP3 over various two-tone spacing, IP3 tends to varies considerably

04-09-98

SGL/EE/ICU

39

MOS Double Balanced UP Converter (1)


s

Peter R. Kinght , IEEE J. Solid-State Circuits, Mar. 1997

04-09-98

SGL/EE/ICU

40

MOS Double Balanced UP Converter (2)


s

Marc A. F. Borremans , IEEE J. Solid-State Circuits, Mar. 1998

04-09-98

SGL/EE/ICU

41

MOS Double Balanced Down Converter (1)


s

G.Giustolisi , IEEE , 1998

04-09-98

SGL/EE/ICU

42

MOS Double Balanced Down Converter (2)


s

Jacques C. Rudell , IEEE J. Solid-State Circuits, Dec. 1997

Arvin R. Shahani , IEEE J. SolidState Circuits, Dec. 1997

04-09-98

SGL/EE/ICU

43

MOS Double Balanced Down Converter (3)


s

Andrew N. Karanicolas, IEEE J. Solid-State Circuits, Dec. 1996

Jan Crols , IEEE J. Solid-State Circuits, July 1995

04-09-98

SGL/EE/ICU

44

MOS Double Balanced Down Converter (4)


s

Seungwook Lee , IEEE, 1997

Ahmadreza Rofougaran , IEEE J. Solid-State Circuits, July 1996

04-09-98

SGL/EE/ICU

45

MOS Single Balanced Down Converter


s

H. Wang, IEEE ISSCC, Feb. 1998

04-09-98

SGL/EE/ICU

46

BJT Double Balanced UP Converter(1)


s

W. Durr , Electronics Letters, Oct. 1998 (  : 5.7GHz,  : 11.2 GHz )

04-09-98

SGL/EE/ICU

47

BJT Double Balanced UP Converter(2)


s

P. Weger , Electronics Letters, Mar. 1991

04-09-98

SGL/EE/ICU

48

BJT Double Balanced UP Converter(3)


s

Farbod Behbahani , IEEE J. Solid-State Circuits, Sep. 1997

Kevin J. Negus , IEEE Transactions on Microwave Theory and Techniques, Sep. 1990

04-09-98

SGL/EE/ICU

49

BJT Double Balanced Down Converter(1)


s

Tsuneo Tsukahara , IEEE J. Solid-State Circuits, Feb. 1996

Jose A. Macedo , IEEE J. SolidState Circuits, Mar. 1998

04-09-98

SGL/EE/ICU

50

BJT Double Balanced Down Converter(2)


s

Barrie Gilbert, IEEE J. Solid-State Circuits, Sep. 1997

Hiroshi Komurasaki , IEEE J. Solid-State Circuits, May 1998

04-09-98

SGL/EE/ICU

51

BJT Double Balanced Down Converter(3)


s

Jeff Durec, IEEE J. Solid-State Circuits, Sep. 1998

04-09-98

SGL/EE/ICU

52

BJT Double Balanced Down Converter(4)


s

John R. Long , IEEE ISSCC, Feb. 1995

Jon Strange , IEEE RF IC Symp. 1997

04-09-98

SGL/EE/ICU

53

BJT Double Balanced Down Converter(5)


s

Robert G. Meyer , IEEE ISSCC, Feb. 1997

Takafumi Yarnaji , IEEE CICC, 1997

04-09-98

SGL/EE/ICU

54

BJT Double Balanced Down Converter(6)


s

Gerhard Schultes , Vehicular Technology Conference Proceedings, May 1990

Jim Wholey , IEEE MTT-S Digest, 1989

04-09-98

SGL/EE/ICU

55

BJT Double Balanced Down Converter(7)


s

Taiwa Lianobu , IEEE Transactions on Consumer Electronics, Aug. 1992

04-09-98

SGL/EE/ICU

56

BJT Double Balanced Down Converter(8)


s

Hisayasu Sato , IEEE J. SolidState Circuits, Dec. 1996

Geoffrey C. Dawe , IEEE ISSCC, Feb. 1997

04-09-98

SGL/EE/ICU

57

BJT Double Balanced Down Converter(9)


s

Werner Simburger , Reprinted with permission from European Microwave Conference, 1995

04-09-98

SGL/EE/ICU

58

BJT Single Balanced Down Converter(1)


s

Robert G. Meyer , IEEE J. SolidState Circuits, Mar. 1994

Volker Thomas , IEEE ISSCC, Feb. 1994

04-09-98

SGL/EE/ICU

59

BJT Single Balanced Down Converter(2)


s

Keng Leong Fong , IEEE J. Solid-State Circuits, Aug. 1997

04-09-98

SGL/EE/ICU

60

InP HBT Double Balanced UP Converter


s

Kevin W. Kobayashi , IEEE J. Solid-State Circuits, Oct. 1994

K.W. Lobayashi , IEEE GaAs IC Symp., 1993

04-09-98

SGL/EE/ICU

61

GaAs UP / Down Converter


s

Yuhki Imai , IEEE Transactions on Microwave Theory and Techniques, Feb. 1991
- MESFET Unbalanced Down Converter

A. Y. Umeda , IEEE Microwave and Millimeter-wave Monolithic Circuits Symp. 1990


- HBT Double Balanced UP Converter

04-09-98

SGL/EE/ICU

62

You might also like