cmos devices and rf building blocks - 東京工業大学 … and circuits...2004.11.11 titech, a....
TRANSCRIPT
2004.11.11 Titech, A. Matsuzawa 1
CMOS Devices and RF Building Blocks
Akira Matsuzawa
Tokyo Institute of Technology
Design of CMOS RF Circuits
2004.11.11 Titech, A. Matsuzawa 2
Contents
• Building blocks in RF systemand basic performances
• MOS Device characteristics in RF application• Basic RF CMOS circuit design
– Low noise amplifier– Mixer– Oscillator
• Future device trend• Summary
2004.11.11 Titech, A. Matsuzawa 3
Current status of RF CMOS chip
• Current products– Bluetooth: 2.4GHz, CSR etc., major– Wireless LAN: 5GHz, Atheros etc., major– CDMA : 0.9GHz-1.9GHz, Qualcomm, becomes major– Zigbee: 2.4GHz, not yet, however must use CMOS– TAG: 2.4GHz, Hitachi etc., major
Major Cellular phone standard uses SiGe-BiCMOS technology
RF CMOS was an university research theme, however currently becomes a major technology in the world of wireless.
2004.11.11 Titech, A. Matsuzawa 4
Why CMOS?
• Low cost– Must be the biggest motivation– CMOS is 30-40% lower than Bi-CMOS
• High level system integration– CMOS is one or two generation advanced– CMOS can realize the full system integration
• Stable supply and multi-foundries– Fabs for SiGe-BiCMOS are very limited.
Slow price decrease and limited product capability• Easy to use
– Universities and start-up companies can use CMOS with low usage fee, but SiGe is difficult to use such programs.
2004.11.11 Titech, A. Matsuzawa 5
Building blocks in RF systemand basic performances
2004.11.11 Titech, A. Matsuzawa 6
Basic RF circuit block
Receiver
Transmitter
ImpedanceMatching
1)LowNoiseAmp. 2) Mixer
3) Oscillator
PowerAmp.
Filter
RF systems are composed of limited circuits blocks.LNA, Mixer, and Oscillator will be discussed in my talk.
2004.11.11 Titech, A. Matsuzawa 7
Basic functions of RF building blocks
Log (f) Log (f)
dB dB1) Amplifier3) Filter
Down conversion
Up conversion
2) Mixer+ Oscillator
Desired
Undesired
Amplifier, frequency converter (mixer +oscillator), and filerare basic function blocks in RF system.
Frequency conversion
2004.11.11 Titech, A. Matsuzawa 8
RF Amplifier
• Gain: Amplify small signal or generate large signal.• Noise: Smaller noise and larger SNR.• Linearity: Smaller non-linearity.
Non-linearity generates undesired frequency components.
.....)()()()( +++= tvtvtvtv inininout3
32
21 ααα
( ) ( )( ) ( ) ( ) ( )( ) ( )( )tttttt 2121212
21 222 ωωωωωωωω ++−+++=+ coscoscoscoscoscos
( ) ( )( ) ( )( ) ( )( ) ....coscoscoscos +−+−=+ tttt 12213
21 2212
21 ωωωωωω
2004.11.11 Titech, A. Matsuzawa 9
Input and output characteristics
Pout
Pin
OIP3
IIP3
IP3
1dBPout(1dB)
Fundamental IMD3
Slope=1Slope=3
SNR minSNR min
SFDRBDR
NoiseFloor
NoiseFloor
SFDR
MDSCP1dB
Distortion and noise are important factors in RF amplifier, as well as power and gain.
2004.11.11 Titech, A. Matsuzawa 10
Dynamic range
BWNFdBmFloorNoise log10174 ++−=
( ) minSNRFloorNoiseIIPSFDR −−= 332
SFDR: Spurious free dynamic rangeThe input power range over which third order inter-modulation productsare below the minimum detectable signal level.
BDR: Blocking dynamic range
kT limitation Bandwidth
MDS: Minimum detectable signal level= Noise Floor +SNRmin
minSNRFloorNoisePBDR dB −−= 1
2004.11.11 Titech, A. Matsuzawa 11
InO
ut
)()()()( txtxtxty 33
221 ααα ++≈
x(t)=Acosωt
( ) ( )
tAtAtAAA
ttAtAtA
tAtAtAty
ωαωαωααα
ωωαωαωα
ωαωαωα
34
224
32
334
212
33
22
33
1
22
33
22
1
333
2221
coscoscos
coscoscoscos
coscoscos)(
++⎟⎟⎠
⎞⎜⎜⎝
⎛++=
++++=
++=
Important
OutputInput
ω 2ω 3ωω
harmonicsFundamental
FundamentalIf a amplifier has a distortion…
y(t)=cosωtcos2ωtcos3ωt
Distortion
Undesired signals are generated by the amplifier’s distortion.
2004.11.11 Titech, A. Matsuzawa 12
20logAin
20lo
gAou
t
1dB
A1-dB
3
11
12131
1450
1204320
αα
ααα
.
loglog
=∴
−=+
−
−
dB
dB
A
dBA 03
In a front end amp.
1-dB compression point
<α
A gain of amplifier will be saturated at 1 dB compression point due to the non-linearity.
)(,
)(
ppmm mVmVVRVP
mWdBmWdBm
62312
101020
2
=∴=
==− μ
2004.11.11 Titech, A. Matsuzawa 13
In
Out
)()()()( txtxtxty 33
221 ααα ++≈
x(t)=A1cosω1t+A2cosω2t
In
322113
22211222111 )coscos()coscos()coscos()( tAtAtAtAtAtAty ωωαωωαωωα +++++=
[ ]
[ ]
[ ]ttAA
ttAAttAA
)2cos()2cos(4
3:2
)2cos()2cos(4
3:2
)cos()cos(:
12121
223
12
21212
213
21
212121221
ωωωωαωω
ωωωωαωω
ωωωωαωωω
−++±=
−++±=
−++±=
ω1ω2 ω1ω2
2ω1-ω2
2ω2-ω1
Out
IntermodulationThe distortion causes undesired signal components around the carrier signals.
Undesired signalsOriginal carriers
2004.11.11 Titech, A. Matsuzawa 14
BPF(Band pass filter)
(BPF)
LNA
ω
ω
ω
2ω1-ω2
ω1 ω2
2ω2-ω1
Signals caused by the distortionfrequency
An amplifier generates undesired signals
2004.11.11 Titech, A. Matsuzawa 15
)log( A120 α
⎟⎠⎞
⎜⎝⎛ 3
34320 Aαlog
inAlog20
outAlog20
Input signal
Out
put s
igna
l
Third intercept point (IP3)
Fundamentalsignal
Third distortionsignal
IIP3
OIP3
IP3
The intercept point, IP3 shows the non-linearity characteristics of amplifier.
3133
13 3
4 IIPOIPIIP ααα
== ,
The larger IP3 means the smaller distortion
2004.11.11 Titech, A. Matsuzawa 16
MOS Device characteristicsin RF application
2004.11.11 Titech, A. Matsuzawa 17
MOS transistor
Gate
Drain
Source
G D
S, B
rg
Cgs
Cgd
gmvg’
vg’ rds
Cds
MOS Transistor Equivalent Circuit
Body
Intrinsic gate voltage and gm are the most important factors in RF CMOS.
2004.11.11 Titech, A. Matsuzawa 18
Cutoff frequency: fT
Cin gmViVi
in
mT C
gfπ2
=∴
Ii
Ii Io
)cos(
)sin(
tCiv
tii
in
ai
ai
ωω
ω
=
=
in
amimo C
tigvgiω
ω )cos(==
fT: Frequency at which the current gain is unity.
Input current
Gate voltageS
GD
Output current
Proportional to gmInversely proportional to Cin
For higher fT, increase gm and decrease Cin.
2004.11.11 Titech, A. Matsuzawa 19
Amplifier gain
insg Cr
1=ω
s
T
rrG 0⋅≈
ωω
in
mT C
g=ω
Cingmvg
Vg
rsvs
ro
ig id Log (G)
Log (f)
G=gmr0
1
sT
ins
om
rr
Crrg 0
0 ωω ==
For the larger gain
Fundamentally larger gmr0 oeff
dsom r
VIrgG ⋅
⎟⎟⎠
⎞⎜⎜⎝
⎛≈≈
2
Larger Ids or ro
Larger Q
CQLQro0
0 ωω ==Q
Distortion and Cin increaseVeff is difficult to reduce
Higher fT and lower rs
For higher voltage gain, increase gm, fT, ro (Q), and decrease input and gate resistance
2004.11.11 Titech, A. Matsuzawa 20
Vgs-Ids特性
W/L=20um/2um0.4um NMOS
( ) )()( dsdsgsds VVLWC
VVVLWC
Ieff
OX
T
OX λμ
λμ
+⎟⎠⎞
⎜⎝⎛=+−⎟
⎠⎞
⎜⎝⎛= 1
21
222
MOSの基本電圧・電流特性: 2乗特性
2004.11.11 Titech, A. Matsuzawa 21
Characteristics of gm (Basic)
⎟⎟⎠
⎞⎜⎜⎝
⎛=
⎟⎟⎠
⎞⎜⎜⎝
⎛=
2
1
2effds
m
eff
dsm VI
gVIg ,
dsox
eff IWL
CV ⋅⋅=
μ1
( )
dsm
effgs
m
gsds
ILWCg
VLWC
dVdI
g
VLWC
VVLWC
I
OX
OX
ds
eff
OX
T
OX
⎟⎠⎞
⎜⎝⎛=
⎟⎠⎞
⎜⎝⎛=≡
⎟⎠⎞
⎜⎝⎛=−⎟
⎠⎞
⎜⎝⎛=
μ
μ
μμ
2
2222
Square law region
Gm is proportional to Ids and inversely proportional to Veff.
Veff is proportional to square root of Ids and inversely proportional to square root of (W/L) ratio.
dsdseff JLWILV ⋅=∝
Veff is proportional to square root of drain current density.
2004.11.11 Titech, A. Matsuzawa 22
サブスレッショルド領域
VgsがVTHよりも低い電圧では拡散により電流が流れる
⎟⎟⎠
⎞⎜⎜⎝
⎛ +−=
⎭⎬⎫
⎩⎨⎧
⎟⎟⎠
⎞⎜⎜⎝
⎛ −−⎟⎟
⎠
⎞⎜⎜⎝
⎛=
T
dsTOTOXso
T
ds
T
gssods
nUVηVexpU
LWCμnI
UVexp
nUVexpII
22
1
qkTUT = 温度電圧:26mV (常温)
η DIBL (Drain Induced Barrier Lowering)係数
)CC(nox
d+= 1 Cd: 空乏層容量
Log(Ids)
VgsVTH0
弱反転領域 強反転領域
ゲート電圧がしきい値電圧よりも低い電圧ではサブスレッショルド電流が流れる。電流はゲート電圧に対して指数関数的に流れる。
サブスレッショルドリーク電流
拡散電流 ドリフト電流
2004.11.11 Titech, A. Matsuzawa 23
速度飽和領域
N+ N+
垂直電界
水平電界
1) Vgが高くなるとチャネルの垂直電界が高くなり表面散乱が大きくなって
モビリティーが低下する。
2) 飽和領域では 水平電界が強くなるとキャリア速度が飽和してモビリティーが劣化する、速度飽和によって電流は飽和する。
1)垂直電界効果 ( )THgseff VV −+
=θ
μμ1
0
2)水平電界効果 o
ss,sos
s
o
μvEEμv,
EEEμ)E(v 2
21
1==
+=
( )s
THgso
THgso
LvVVμ
VVLμv
21 −+
−=( )vVVWCvQI THgsox
ds 2−
=⋅=
垂直電界によるモビリティーの劣化効果を入れて
( )( ) LvVVVV
LWCI
s
o
Tgs
THgsoxods 212
2 μθξξ
μ+=
−+−
⎟⎠⎞
⎜⎝⎛≈ ,
2004.11.11 Titech, A. Matsuzawa 24
ドレイン電圧 VD
ドレイン電
流I D
(VG-VT)2に比例して増加する
ドレイン電圧 VD
ドレイン電
流I D
ゲート電圧に対してほぼ等間隔
傾斜
)VV(WCvIMOSFET
)VV(CμLWI
MOSFET
THgsoxSds
THgsoxods
−=
−=
21
2
短チャネル
長チャネル
実際はこの中間を取り、以下の表現を用いる場合もある。
微細トランジスタの電圧・電流特性
( )αTgsoxds VVCμLWI −= 0
21
α: 1~2, 通常1.3程度桜井のα乗則T. Sakurai, et al., IEEE, JSC, Vol. 25, no.2, pp.584-594, 1990.
微細なトランジスタではゲート電圧に比例する電流になる。
2004.11.11 Titech, A. Matsuzawa 25
Non-ideal effects to square low region
Tds
m
T
dsm
gssods
nUIg
nUIg
nUV
IIT
1==
⎟⎟⎠
⎞⎜⎜⎝
⎛≈ exp
Mobility degradation and velocity saturation
Low Veff Sub-threshold region
High Veff
LvV ceff
00
0
1μθθ
θμμ +≈
+≈ ,
( Weak inversion)
This effect becomes larger at large Veff and short channel length.
At larger Veff and lower Veff, two non-ideal effects are not negligible .
2004.11.11 Titech, A. Matsuzawa 26
Distortion
⋅⋅⋅⋅+++= 33
221 effeffeffds VaVaVaI
3
133
3
3 34
61
aaIIP
dVIdaeff
ds =≡
Lower Veff gives higher gm, bur results in higher distortion.To obtain lower distortion ( higher IIP3), we must increase Veff.
Higher gm and lower distortion means higher Ids.
2004.11.11 Titech, A. Matsuzawa 27
速度飽和効果と歪
Lv
VVV
LWKI
sat
effdseffds
2
11
11
0
2
μθ
λ
+=Θ
Θ+−=
( )( )
( )( )23
2
123
4
12
effeffeff
effeffeff
VVV
IIP
VVVIIP
Θ+Θ+Θ
=
Θ+Θ+=
・Veffが大きいほど歪が小さい・微細化により歪は小さくなる(速度飽和効果)
2004.11.11 Titech, A. Matsuzawa 28
ドレイン抵抗
ドレイン抵抗はVdsがVeffよりも大きくなればリニア領域から飽和領域に入って一定値を取るもの
ではなく、ドレイン電圧により大きく変化する。Vdsが低い領域ではドレイン抵抗が小さく、低電圧で高利得の回路を設計する際注意を要する。
2004.11.11 Titech, A. Matsuzawa 29
飽和領域のドレインコンダクタンス
xxdsds WkIg −= 1
0.4umCMOSの実測
gdsを下げる(rdsを上げる)ためにはゲート長Lを大きくする。Lminの場合はDIBL効果によりrdsが低いので注意が必要。
Lgds
1∝
飽和領域のドレインコンダクタンスgdsはチャネル長に反比例する。
2004.11.11 Titech, A. Matsuzawa 30
ピーク遮断周波数
速度飽和領域でのgm
ピークの遮断周波数はチャネル長に反比例する(0.1umでは120GHz程度に達する)
チャネル長で決定
LWCg
Cgf
ox
m
gs
mT ππ 22
≈≈
satoxmsat vWCg =
Lvf sat
Tpeak π2=∴
2004.11.11 Titech, A. Matsuzawa 31
GHz operation by CMOS
in
mT C
gfπ2
≡
Cutoff frequency of MOS becomes higher than that of Bipolar.Over several GHz operations have attained in CMOS technology
eff
satTpeak L
vfπ2
≈
2004.11.11 Titech, A. Matsuzawa 32
Effect of parasitic capacitance to fT
)( pgdgs
mT CCC
gf++
≡π2
fT of actual circuit is reduced by the parasitic capacitance.There is an optimum gate width to obtain highest fT.
Cingmvi
ViCp
Cin=Cgs+Cgd
Region(1); Increased by increasing gmRegion(2); Decreased by increasing Cin
2004.11.11 Titech, A. Matsuzawa 33
fT: MOS vs. Bipolar
⎟⎟⎠
⎞⎜⎜⎝
⎛≡
2eff
dsm V
IgT
cm U
Ig ≡
Teff nUV 2=min n: 1.4 BipCMOS gmgm41,
21
<mV
qkTUT 26≈≡
(Same operating current)
in
mT C
gfπ2
≡
BipCMOS CinCin41,
21
<
(Same fT)
Veff/2: 50-100mV(actual ckt.)
Even if fT of MOS is same as that of Bipolar, fT of MOS is easily lowered by parasitic capacitance.Because, gm of MOS is ½ to ¼ of that of Bipolar at the same current.
MOS Bipolar
Small parasitic capacitance is a key for RF CMOS design.
2004.11.11 Titech, A. Matsuzawa 34
VT mismatch
VT mismatch degrades accuracy; ADC, OP amp, and Mixer.Larger gate area is needed for small VT mismatch.Scaling and proper channel structure improves mismatch.
2004.11.11 Titech, A. Matsuzawa 35
RF noise source in MOS transistor
N+ N+
Gate resistance noise
md kTBgi382 =
Induced Gate Noise
2. Channel noise
1. Gate resistance
Channel noise
3. Induced Gate Noise gg kTBgi 42 =do
gsg g
Cg
22
51 ω
=
Ig2 and id2 has a strong correlation
(C=0.4j)
dod gkTBior γ42 =,
gn kTBRe 42 =
gdo: zero-bias channel resistance in linear region
Conventional SPICE does not include this noise
mn g
kTBv3
82 =
2004.11.11 Titech, A. Matsuzawa 36
1/f noise1/f noise of MOS is larger than that of bipolar.
22oxvf
vfnf TS
ff
LWS
V ∝Δ
= ,
For the lower 1/f noise, the larger gate area is needed.
2004.11.11 Titech, A. Matsuzawa 37
Substrate effect
SDSDSPAD
Gate
SDSDSPAD
GateShield layer
RF power loss and noise generation
Substrate should be treated as resistive network.
This substrate resistance causes RF power loss and noise generation.
Shielding can reduce this effect.
2004.11.11 Titech, A. Matsuzawa 38
Power loss in substrate
C
RpGp Cp
Equivalent
CR
CC
RG
pp
p
p
p
p
pp
1
1
1
1
1
2
2
2
=
⎟⎟⎠
⎞⎜⎜⎝
⎛+
=
⎟⎟⎠
⎞⎜⎜⎝
⎛+
⎟⎟⎠
⎞⎜⎜⎝
⎛
⋅=
ϖ
ωω
ωω
ωω
Very low resistance or high resistance realizes low power loss.
Higher C and moderate Rsubresults in higher power loss.
2004.11.11 Titech, A. Matsuzawa 39
LC resonator
CQLQro
00
ωω ==Substrate
LC
r0
Q
LC1
0 =ω
LC resonator can be regarded as resistance at the resonance frequency.
2004.11.11 Titech, A. Matsuzawa 40
インダクタ
2
1Q
S ∝φ
RF-CMOS LSI
QI 1∝
Noise:
Current
L
C
L
CVc
Vb
Vo Vo
RF-VCO
VCOの性能の大半はインダクタのQで決定される。したがってきわめて重要な素子であるが、同時に最も面積を占有する素子である。最近のRFLSIはインダクタをできるだけ用いないようにしている。
2004.11.11 Titech, A. Matsuzawa 41
インダクタの形状とインダクタンス
2004.11.11 Titech, A. Matsuzawa 42
インダクタのQ
3層メタル(厚さ 3um)基板抵抗 500Ω
低い周波数では配線抵抗が支配的高い周波数では寄生容量を通じて発生するロスが支配的
配線膜圧を厚くする
上層メタルを用いる高比抵抗基板を用いる
2004.11.11 Titech, A. Matsuzawa 43~ 0.35-μm CMOS /SOI Devices ~
高比抵抗基板の効果:インダクタのQの向上
高比抵抗基板では高周波信号の基板での損失が小さいためにインダクタのQが向上する。
2004.11.11 Titech, A. Matsuzawa 44
パターンシールド
誘導電流による損失を防ぐためにインダクタと基板間にポリシリコンなどでパタンシールドを挿入することが行われている。誘起されるループ電流に対して直角にスリットを入れるか、トレンチの溝を切る
2004.11.11 Titech, A. Matsuzawa 45
バラクタ
MOSバラクタは通常、蓄積型MOS容量が用いられる。Qが高く100程度のものが多い。
2004.11.11 Titech, A. Matsuzawa 46
Basic RF CMOS circuit design
Low noise amplifier
Mixer
Oscillator
2004.11.11 Titech, A. Matsuzawa 47
Lg
Ls
Cpi
Rsub
M1
M2
Z0 rg
Cgs rgs
• High gain• Impedance matching• Low noise• High IIP3
• Low noise
Conventional MOS LNA
Low noise amplifier design
2004.11.11 Titech, A. Matsuzawa 48
Noise figure: General
sss jXRZ +=
( )nis
s
nv
s
nis
s
nv
rsn
ngsngrsn GRRR
RGR
RR
V
IRVVF ++≈++=
++= 11
2
2
22
,
,
ningnvng kTGIkTRV 44 22 == ,
NoiselessCircuit
Vs
Rs Vn,rs Vng
Ing
The lower Rnv and Gni realizes the better for lower noise figure.
ni
nv
ng
ngsopt
GR
IVR == ninvGRF 21+≈min
Noise voltage source
Noise current source
Described by equivalent resistance and conductance
231NL
WRr totsrg =
( )ds
eff
mNQSgs
gsgnv
IV
gr
rrR
1051
≈≈
+=
,
2
0
5 ⎟⎟⎠
⎞⎜⎜⎝
⎛=
T
mni
gGωω
2004.11.11 Titech, A. Matsuzawa 49
Low NF design
( )
)(
.
0
20
51
1
4051
1
ωω
ωω
>>
++≈
+⎟⎠⎞
⎜⎝⎛+
++≈
Ts
mg
smTs
mg
rg
r
rgrg
rF
S
D
S
D
S
S
D
Divide the gate
231NL
WRr totsrg =
(いろいろ調べたが、まだ定説になっている理論式は無いと言ってよい。微妙に異なっている。)
Rsr: Sheet resistanceN:The # of division
Low noise figure
1) Lower the gate resistance
Dived the gate or lower the gate sheet resistance
2) Reduce substrate loss
Use shield technique to the input bonding PAD.
Use high resistive substrate, if possible.
3) Increase drain current
4) Increase rs, if possible.
Reduce parasitic capacitance
2004.11.11 Titech, A. Matsuzawa 50
( )
( )
( )
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛−+=→
++=≡∴
−+=
=+−+
+
+
gsgs
gs
sin
gsgs
gs
sm
t
tin
sstmts
ttgs
sstmt
CLLj
CLg
Z
sCLLs
CLg
ivZ
sLvvgiv
visC
sLvvgi
m
ωω 1
1
1
)(
)(
Vin
Ls
vt
it vs
( )stm vvg −
Impedance matching
Lg
Cgs Ls
入力インピーダンスに合わせる
ゼロにする
入力インピーダンス整合はこのような回路により実現できる
2004.11.11 Titech, A. Matsuzawa 51
Ids and Veff optimization
WIV ds
eff ∝
sdseffsmLNA
LNA rIVrgFIIPDR ≈∝
−∝
13
Veff
NF
Gain
3rd
distortion
dB
Adjust the Ids and Veff for optimization of gain, noise and distortion.
Lower Ids
effVIIP
Higher Ids
∝3
Dynamic range of LNA is proportional to Ids.
L
moi CgQQGainω
≈
動作電流の設定
Veff の設定
NF, IP3, Gain整合などのチェック
OK?YES
NO
大きいほど高性能
高:gm小、fT高、歪小低: gm小、fT低、歪大
Veff 高, W小Veff 小, W大
2004.11.11 Titech, A. Matsuzawa 52
チャネル長の選定
LW
LLWVT ≡=∝Δ α
αQ
11
31.LS =
ds
OXVT
eff
T
ds
ds
nIC
LVV
II μσ 22
≈Δ
=Δ
Lro ≈
LLWVn
11≈∝
Log L 大きい小さい
Log
大きさ
VTばらつき
トランジスタの面積
電流源ばらつき
出力抵抗
1/fノイズ
ゲート容量2LLWCC oxgs ∝=
DCゲイン LrgG om ∝=
周波数特性 LGCLCCg
gdoxp
mT ++
∝ 2ω
Lにより性能が変わる。
さまざまな考慮で最適点を決めていく。
2004.11.11 Titech, A. Matsuzawa 53
NF progress in MOS LNA
NF of MOS LNA is reaching 1dB.
2004.11.11 Titech, A. Matsuzawa 54
Vout
VDD
LPF
Amp 1
Amp 2
DC Bias current
RF pathTwo stage amp. Reuse of Bias current
A.R.Shahani, ISSCC Dig. Tech. Papers, p.368 (1997)
HPF
Circuit example
同一電流で2段の増幅器になる。
2004.11.11 Titech, A. Matsuzawa 55
Mixer
( )( )tAV LOsso ωωπ
±= cos2Vs
VLO
Vo( )tAV sss ωcos=
( )tAV LOLOLO ωcos=
If VLO>>4Veff (Full swing)
FIF FIF
Freq
dB
FIF
Freq
dB
FLO
Fdes
Fimage
RF spectrum IF spectrum
Mixer converts frequency, but image signal is converted to the same frequency.
2004.11.11 Titech, A. Matsuzawa 56
Image-reject mixers
( )tLOωcos
LPF
+
LPF °− 45
Vin (t) Vout(t)
( )tLOωsin
°45
( ) ( )tAtAtV imimdesdesin ωω coscos)( +=
( ) ( )tIAAtAAtV IFRcimIFcdesout ωω coscos)( +=
Ac: Conversion gain, IR: Image rejection
IR=0 if I/Q phase difference is 90° and Channel conversion gains are equal.
Quadrature mixing realizes image-suppression.Gain and phase matching is needed.
2004.11.11 Titech, A. Matsuzawa 57
Gain mismatch and phase error
φγγφγγ
coscos
2121
2
2
++−+
=desired
spur
PP γ
φ: Gain ratio
:Phase error
A. Rofougaran, et al., IEEE J.S.C. Vol.33, No.4, April 1998. PP. 515-534.
2004.11.11 Titech, A. Matsuzawa 58
Passive FET mixer
Lo Lo
LoLo
Vin
Vin
Vo Vo
Low powerHigh linearityNo 1/F noise
No conversion gainNo isolation, Bi-directional
Passive FET mixer
MOS can realize passive mixer easily.
Ultimately low power, but take care of isolation.
2004.11.11 Titech, A. Matsuzawa 59
Active mixers
Lo Lo
Vo Vo
Vin
RL
Single balanced mixer
Lo Lo
Vo
Vin
Lo
Vin
Vo
Double balanced mixer
M1
M2 M3
M1 M1’
M2 M3 M2’ M3’
Very small direct feed through and even order distortion
RL RL RL
Direct LO component
2004.11.11 Titech, A. Matsuzawa 60
Active Mixer design
Lo Lo
Vo
Vin
Lo
Vin
Vo
M1 M1’
M2 M3 M2’ M3’
RL RL
RF amplifier
Switch
Load
Low noiseLow distortion
High gainSame as LNA
Noise: Thermal and 1/f
Gain and noise
2004.11.11 Titech, A. Matsuzawa 61
Active mixer design
Lmmix RgG 12π
=
eff
ds
eff
mLm
onin
mLLmLO
LLon
VIIP
IV
kTg
kTRg
vSSBv
gkTRRgAIRkTRv
≈
≅≈
⎟⎠⎞
⎜⎝⎛
=
≈⎟⎟⎠
⎞⎜⎜⎝
⎛++=
3
2
1
22
1
22
12
12
22
8218
γπγπ
π
γγπγ
Mixer gain
Thermal noise
A larger dynamic range needs larger current
Ts TLO
1/F noise
1) Switch transistor (M2, M3)
gsCWLswn
swnLO
son
v
vTTv
112
4
∝≈
=
,
,,
Larger Ids is needed for higher dynamic rangeand shorter switching time for low 1/f noise.
2) Load transistors Directly produces
Phase modulation
Shorter switching time or larger Ts/TLo ratio
Load resistor
Input transistor
LO swing
Noise in switch
2004.11.11 Titech, A. Matsuzawa 62
1/f noise
1/f noise of MOS is larger than that of bipolar.
22 , oxvfvf
nf TSff
LWSV ∝
Δ=
For the lower 1/f noise, the larger gate area is needed.
2004.11.11 Titech, A. Matsuzawa 63
1/f noise generation
1/f noise can be expressed as small imbalance signal to the differential pair.
This voltage modulates the switching timing.
Filtered (Integrated) pulse train generates 1/f noise on the Load.
H. Darabi and A. Abidi; JSC, vol. 35, No. 1, pp.15-25, Jan. 2000
This effect can be expressed as a narrow current pulse train
1/f noise from mixer circuits causes serious problem in direct conversion and low IF conversion.
2004.11.11 Titech, A. Matsuzawa 64
Principle of Oscillator
L C L C RL C R -R
Damped waveNon-damped wave Non-damped wave
CLQR =LCo
1=ω
mgR 2=−
Negative resistor sustains oscillation
2004.11.11 Titech, A. Matsuzawa 65
CMOS LC Oscillator
M2 M3
L
C
L
C
Vb
Vc
Vo Vo
M1
M2 M3
I
+Vo -Vo
Rin
min gR 2
=−
2004.11.11 Titech, A. Matsuzawa 66
Oscillator
πo
oscIrV 4
=
QCV
rVI odd
o
ddopt
ωππ==
2
CQLQro0
0ω
ω ==
L CQr0-1/gm -1/gm
1) Amplitude condition
Oscillationamplitude
Vdd
2) Oscillation condition
Headroomlimit
2Vdd
QCV
Ir
g effo
om
3232
2 ,,, ,
ω>>
L
C
There is an optimum Ids for low phase noise.
L
C
Vb
Vc
Vo Vo
2004.11.11 Titech, A. Matsuzawa 67
ωc ω ωc ω
Ideal oscillator Actual oscillator
Phase noise in oscillator
( )1<<
−≈+
)(sin)(cos)(cos
tttAtAttA ccc
θωθωθω
Q
Phase noise causes broad noise spectrum around carrier frequency
2004.11.11 Titech, A. Matsuzawa 68
Periodical phase noise
( ) ( )( )
( ) ( )[ ]ttAtA
ttAtAttAttAty
radtt
mcmcm
c
mmcc
mmcmmcn
mmmn
ωωωωφω
ωφωωωφωωφω
φωφφ
−−++=
⋅−≈⋅−⋅=
<<=
coscoscos
sinsincossinsinsinsincoscos)(
.sin)('
2
1
[ ])(cos)( ttAty nc φω +=
tA
ωc
ωc+ωm
ωc-ωm
Side band
Periodical phase modulation causes sidebands
2004.11.11 Titech, A. Matsuzawa 69
Phase noise problem
Oscillator phase noise causes interference to wanted signal.This reduces SNR for received signals.
2004.11.11 Titech, A. Matsuzawa 70
Phase noise of oscillator
2
0222
24 ⎟⎟
⎠
⎞⎜⎜⎝
⎛=⋅
Δ=
Δ m
onn
QkTrZ
fi
fv
ωω
( )
0
00
2ωωωωω
mm
LjZ ≈+
LrQ0
0
ω=
( )ωjZ
0
21ωQ
BW=
BW
ω
R
0.7R
L CQ
-1/gm -1/gmr0
( )m
m QrZωωωω
200
0 ≈+
{ }⎥⎥⎦
⎤
⎢⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛⋅=
2
0
2210
msigm QP
kTLωωω log
0ωω <<m
Noise spectrum density
Phase noise
(Filter action)
2004.11.11 Titech, A. Matsuzawa 71
Phase noise equation
{ }⎥⎥⎦
⎤
⎢⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛⋅⋅=
2
0
2210
msigm QP
kTLωωω log
{ }⎥⎥⎦
⎤
⎢⎢⎣
⎡
⎟⎟
⎠
⎞
⎜⎜
⎝
⎛+
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
⎟⎟⎠
⎞⎜⎜⎝
⎛+⋅⋅=
m
m
msigm
f
QPFkTL
ω
ω
ωωω
3112
12102
0 /log
Larger signal powerHigher Q for lower phase noise
F: Noise factor
Modification
Noise floor Thermal noise filtered by LC tank
1/f noise effect
Leeson’s equation
D. B. Leeson, Proceedings of the IEEE, vol. 54, pp. 329-330, Feb. 1966.
Due to the filter action of LC tank.
2004.11.11 Titech, A. Matsuzawa 72
Frequency characteristics of Phase noise in oscillator
0
2ω
LQBW =
3
2
0 122 msL P
FkTQ ωω
⎟⎟⎠
⎞⎜⎜⎝
⎛
2
2
0 122 msL P
FkTQ ωω
⎟⎟⎠
⎞⎜⎜⎝
⎛
sPFkT2
coω
-9dB/oct(Slope =-3)
-6dB/oct(Slope=-2)
m
Phas
e no
ise
spec
trum
ω
1/f noise Thermal
( )
Thermal
1/f noise and thermal noise is converted to 1/f3 and1/f2, respectively.
mm
aSω
ωθ =Δ
( )PsFkTS m
2=Δ ωθ
( )mmL
m SQ
S ωωωω θφ Δ⎟⎟
⎠
⎞⎜⎜⎝
⎛=
2
0
21)(
2004.11.11 Titech, A. Matsuzawa 73
Up and down converted noise
)/( HzVVnoise
ωo 2ωo 3ωo
P (dBm)
ω
ω
Down-conv.Up-conv.
Noises around 2fo are down converted to fo.
Noise into tank
2004.11.11 Titech, A. Matsuzawa 74
FoM and minimum phase noise
IVfLffF
ddmmoM )(
12
0⎟⎟⎠
⎞⎜⎜⎝
⎛=
⎟⎟⎠
⎞⎜⎜⎝
⎛⋅⎟⎟
⎠
⎞⎜⎜⎝
⎛⋅⋅=⋅⎟⎟
⎠
⎞⎜⎜⎝
⎛⋅⋅=
o
om
o
RFm
om
rVFkT
ff
QPFkT
ff
QfL
2
1211
21
2
2
2
2
2)(
19882 mo
o
o grVIrF ⋅++= γ
πγ
Fm: Offset frequencyL(fm): Phase noise at offset freq.
F: Noise factor
indo
ddodd
o
ddopt LQ
VQ
CVrVI
ωπωππ
22===
2
1
2
93242
14 Q
VVkT
QF
eff
ddoM ∝
++=
,
γπγπ
FoM is basically proportional to Q2.
at Iopt
Tank loss Switch noise Current source
2004.11.11 Titech, A. Matsuzawa 75
Optimization
Bias currentIopt
Phase noise Oscillationamplitude
2Vdd
Large VddLarge Veff1, but take care of Vo reductionLarge L1, W1 to reduce 1/f noiseProper W/L for M2, M3Higher QLarger QLind for Lower Iopt
2
1
212 ⎟⎟
⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛+⋅⋅⋅=
m
o
effdd
indo
ddm f
fVVQ
LV
kTfL,
min )( ωγ
Careful optimization reduces oscillator phase noise.
L
C
L
C
Vb
Vc
Vo Vo
M1: Veff1, L1, W1
M2 M3
Small W2, W3 for higher amplitude and low CpSufficient W2, W3 for acceptable gm
2004.11.11 Titech, A. Matsuzawa 76
L
C
L
CVc
Vb
Vo VoL
C
L
C
Vb
Vc
Vo Vo
L
C
L
C
Vb
Vc
Vo Vo
Cx
LxCs
Hi-Z at 2fo
(a) (b)(c)
CMOS oscillator circuits
Basic Low power (gm is higher) Low noise by filtering
E. Hegazi, ISSCC 2001
2004.11.11 Titech, A. Matsuzawa 77
Filtering of 2fo component in OSC.
Noise filtering of 2fo component reduces OSC phase noise to -10dB.
E. Hegazi, ISSCC 2001
2004.11.11 Titech, A. Matsuzawa 78
Oscillator phase noise progress
Phase noise in CMOS oscillator becomes lower than that of bipolar.
2004.11.11 Titech, A. Matsuzawa 79
Future device trend
2004.11.11 Titech, A. Matsuzawa 80
現在のSoC用トランジスタ
現在のSoCの量産プロセスである0.13umルールのトランジスタ
原子レベルの制御が求められる。
松下電器
2004.11.11 Titech, A. Matsuzawa 81
CMOSの微細化と高周波特性の向上
微細化とfTの上昇は今後も続くが、、、
2004.11.11 Titech, A. Matsuzawa 82
微細化と動作電圧
微細化による動作電圧の低下は緩和されたものの今後のCMOSは1V以下の電圧で動作さなければならない。
2004.11.11 Titech, A. Matsuzawa 83
微細化とノイズ
微細化とともに熱雑音係数は増大 微細化とともに1/fノイズは減少するがHi-Kの導入は1/fノイズを増大させる。
2004.11.11 Titech, A. Matsuzawa 84
1/fノイズとVTミスマッチトレンド
・1/fノイズの係数は比較的順調に減少と予測・VTミスマッチはあまり改善されないと予測
C. H. Diaz, et al., IEEE, Tran on ED, Vol. 50, pp.557-566、2003
2004.11.11 Titech, A. Matsuzawa 85
微細デバイスのドレイン抵抗
ds
effA
dsds I
VVg
r+
≈≡1
微細デバイスではポケット注入を用いていることによりチャネル長を伸ばしてもVAつ
まりはドレイン抵抗はあまり上がらない。つまり、微細プロセスではDC利得が極めてあげにくいことを意味する。
D, Buss, et al., IEEE, Tran on ED, Vol. 50, pp.546-556、2003
2004.11.11 Titech, A. Matsuzawa 86
インダクタンストレンド・配線層数の増加や最上層配線の厚膜化などによりインダクタンスのQは増加傾向
C. H. Diaz, et al., IEEE, Tran on ED, Vol. 50, pp.557-566、2003
最近のインダクタは6um程度の厚さを用いて20以上のQを得ているものもある。
2004.11.11 Titech, A. Matsuzawa 87
Cost up issue by analog parts
Cost of mixed A/D LSI will increase when using deep sub-micron device, due to the increase of cost of non-scalable analog parts.
Large analog may be unacceptable.Some analog circuits should be replaced by digital circuits
2004.11.11 Titech, A. Matsuzawa 88
Technology edge RF CMOS LSI
M. Zargari (Atheros), et al., ISSCC 2004, pp.96 K. Muhammad (TI), et al., ISSCC2004, pp.268
Discrete-time Bluetooth0.13um, 1.5V, 2.4GHz
Wireless LAN, 802.11 a/b/g0.25um, 2.5V, 23mm2, 5GHz
Many RF CMOS LSIs have been developed for many standards
2004.11.11 Titech, A. Matsuzawa 89
Feature of CMOS technology
• Pros– Can use switch and voltage controlled conductance– Smaller distortion– No carrier accumulation – Can use switched capacitor circuits– Can increase fT by scaling– Easy use of complementally circuits– Easy integration with digital circuits
• Cons– Low gm/Ids– Larger mismatch voltage and 1/f noise– Lower operating voltage with scaling– Difficult to enable impedance matching– Easily affected by substrate
2004.11.11 Titech, A. Matsuzawa 90
Potential and limitation of RF CMOS
• Potential– RF CMOS are going to be major in wireless products.– SiGe Bi-CMOS technology will be used for extremely low noise and
low power RF products.– 60GHz or 100GHz CMOS circuits has already developed.
• Limitation– For low noise and low power characteristics in RF circuit, SiGe
bipolar technology must be better than CMOS. – High power PA will not be integrated on scaled CMOS chip, due to
low efficiency and needs low voltage operation.– Use of further advanced technology beyond 90nm would be limited.
This is because low analog operating voltage of less than 0.9V and chip and development cost increase.
2004.11.11 Titech, A. Matsuzawa 91
Advantage of SiGe Bi-CMOS
Alcatel1) OkiProcess
Chip size゙
Id
Rx_Sens.
Tx_Power
Rx
Tx
Broadcom Conexant MitsubishiCMOS0.25um
40.1mm2
50mA
-80dBm
+2dBm
70mA
CMOS0.35um
18.0mm2
47mA
-77dBm
+2.5dBm
66mA
CMOS0.35um
46mA
-80dBm
+5dBm
47mA
SiGe-BiCMOS0.5um
12mA
-78dBm
+2dBm
16.4mA
Si-BiCMOS0.5um
34.4mA
44.0mA
-80dBm
0dBm
17.0mm2
Vdd 2.7-3.3V1.8-3.6V3V(?)2.7-3.3V2.5V
SiGe Bip has a great advantage in power consumption
2004.11.11 Titech, A. Matsuzawa 92
CMOS vs. SiGe-BiCMOS
Wafer cost of SiGe BiCMOS is 30-40% higher than CMOS at the same generation,however almost same as one generation advanced CMOS.
SiGe-BiCMOSはコスト高と供給問題があったため一部のメーカーしか使用できなかったが、状況は改善し、0.13um程度の微細化も進んでいるようである。
今後、もっと使用されるものと思われる。
2004.11.11 Titech, A. Matsuzawa 93
Principal design for RF CMOS
• Use small size devices and compensate the accuracy and 1/f noise degradation.– Small parasitic capacitance is imperative.
• Much smaller capacitance is needed to keep higher cutoff frequency under the lower gm condition
– Small size results in increase of mismatch voltage and 1/f noise• Should be addressed by analog or digital compensation,
not by increase of device size.
• Keep the voltage swing high as possible to realize higher SNR– The noise level is higher and gm is lower than those of bipolar.
• Use digital technology rather than analog technology– Performance increase and power and area decrease are promised
by scaling. Analog is not so.
2004.11.11 Titech, A. Matsuzawa 94
Acknowledgment and references• Acknowledgment
I would like to thank Prof. Asad Abidi from UCLA and N. Itoh from Toshiba for their advices.
• References– Asad A. Abidi, “Power-Conscious design of Wireless circuits and systems,” pp.665-695,
“Trade-offs in Analog Circuit Design,” Kluwer Academic Publishers, 2002. (Edited by Chris Toumanzou, George Moschytz, and Barrie Gilbert)
– Thomas. H. Lee, “The design of CMOS RF ICs,” Cambridge University Press, Jan. 1998.– Bezad, Razavi, “RF micro-electronics,” Prentice Hall, Nov. 1999.– Domine Leenaerts, Johan van der Tang, and Ciero Vaucher, “Circuit Design for RF
Transceivers,” Kluwer Academic Publishers, 2001.– Charles Chien, “Digital Radio Systems on A chip,” Kluwer Academic Publishers, 2001.– E. Hegazi, et. Al., ”A Filtering Technique to Lower Oscillator Phase Noise,” ISSCC 2001,
23.4, Feb. 2001.– 伊藤信之, “RF CMOS回路設計技術” トリケップス 2002年– 松澤昭 “CMOSのRF応用” 応用物理, pp.318-322, Vol.71, No.3, 2002.– “Special Issue on Silicon-Based RF and Microwave Integrated Circuits”
IEEE Transaction on Microwave Theory and Techniques, Vol. 50, No. 1, Jan. 2002.
– “Special Issue on Device Integration Technology for Mixed-Signal SoC”IEEE Transaction on Electron Devices, Vol. 50, No. 3, March, 2003.