Upload
others
View
13
Download
0
Embed Size (px)
Citation preview
A 0.7 V-to-1.0 V 10.1 dBm-to-13.2 dBm
60-GHz Power Amplifier Using Digitally-
Assisted LDO Considering HCI Issues
Rui Wu, Yuuki Tsukui, Ryo Minami, Kenichi Okada,
and Akira Matsuzawa
Tokyo Institute of Technology, Japan
1
Outline
• Background
– 60-GHz field is attractive
• Hot-Carrier-Induced Issues
– HCI influence on circuit reliability
• Variable-Supply-Voltage PA using
Digitally-assisted LDO
– Circuit design & Measurement results
• Conclusions
2
Background
[1] http://www.tele.soumu.go.jp
• 9-GHz unlicensed bandwidth
• Several Gbps wireless communication
e.g. IEEE 802.15.3c
QPSK3.5 Gbps/ch
16QAM7 Gbps/ch
3
HCI Issues are Emerging at 60 GHz
High fmax, suitable for
60-GHz amplifier
Bad HCI performance
60-GHz power amplifier
Standard
Good HCI performance
Low fmax, can’t be used
for 60-GHz amplifier
2.4-GHz power amplifier
Thick oxide
Standard
4
Hot-Carrier-Induced (HCI) Effects
P-Substrate
Impact ionization
n+n+Oxide
damage
Gate
Source Drain
Isub
Ig
SiO2
Degrade Vth, gm, drain current, and lifetime
5
1E+00
1E+02
1E+04
1E+06
1E+08
1E+10
0.4 0.5 0.6 0.7 0.8 0.9
65 nm NMOSFET DC Stress Lifetime
1/VDS (1/V)
100
102
104
106
108
1010
Lif
eti
me t
(s
)
[2] E. Takeda et al., IEDL 1983
VGS
VDS
VDS
t
V
Stress condition
VGS = 0.8 V
Lifetime is the time drain current decreases by 10%
6
1
10
100
1E+02 1E+03 1E+04 1E+05
65 nm NMOSFET RF Stress Lifetime ∆
I DS
at (%
)
Freq.=100 MHz,
Po=11 dBm
102 103 104 105
Time (s)
[3] L. Negre et al., JSSC 2012
Vgs
Vds
Stress condition
t
0.8 V
1.2 V
Vds Vgs
7
Hot-Carrier Damage Mechanism [4]
• Single Vibrational Excitation (SVE) is related to
high energetic carrier that has enough energy to
break Si-H bond; (High energy)
• Electron Electron Scattering (EES) is caused by
one carrier promotes the other into higher energy
and allows Si-H breaking; (Medium energy)
• Multiple Vibrational Excitation (MVE) is due to a
series of low energetic carriers that accumulate
enough energy to break Si-H bond. (Low energy)
[4] C. Guerin et al., JAP 2009
8
Hot-Carrier Physical Model [3]
IDS(t) is the time-varying drain current which is a function
of VGS(t) and VDS(t);
K and a are damage mode dependent constants;
m and n are process-related constants.
[3] L. Negre et al., JSSC 2012
9
Conventional Solutions
[6] M. Tanomura et al., ISSCC 2008
[5] A. Siligaris et al., JSSC 2010
[7] J. Chen et al., ISSCC 2011
Better lifetime
Degraded output power,
efficiency, and linearity
Cascode [5]
Low Vdd [6]
Better lifetime, output
power, and linearity
Sensitive to process
variations
Power combining [7]
10
Application Scenario
◆Single Carrier (SC) Mode of IEEE 802.15.3c
MCS identifier 8 1
Data rate 2640 Mb/s 412 Mb/s
Rx sensitivity -56 dBm -61 dBm
Required CNR 17.5 dB 12.5 dB
Distance 0.56 m 1 m 1 m
Required Pout 5 dBm 10 dBm 5 dBm
Mod 1+MCS 8
0 d0.56 m 1 m
Mod 1+MCS 1
&
Mod 2+MCS 8
*NF=8 dB; Thermal noise=-81.5 dBm; Antenna gain=2 dBi;
Implementation loss=-2 dB; freq.=60 GHz
Mod1 low power
Mod2 high power
11
Lifetime estimation
12
Time-Division Duplex (TDD) Operation
SIFS SIFSRX
TX TXACK
...
...
Time
Time
In IEEE 802.11ad, short inter-frame space (SIFS)
is indicated to be 3ms
• TDD operation can eliminate the
stringent requirement of filtering and
extend the available bandwidth for
transceivers.
13
The Proposed Power Amplifier
14
Transient Operation of the LDO
VPA
Time
βVref
VrefD
Training
βV’ref
V’refD
Sleep
RecordRestore
Awaking Awaking
Restore
Record
15
Transient Simulation Result
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 1 2 3 4
0.1 ms
Time (ms)
VP
A (
V)
CLK=200 MHz
CL=86 pF
16
Differential PA Topology
Vdd VPAVddVg1 Vg2 Vg3
M1M2
M3
M4 M5 M6
RFin+
RFin-
RFout-
RFout+
Cc1 Cc2
Cc1 Cc2
MIM TL TL
The cross-coupling capacitor technique is adopted
to improve the stability and power gain
17
Die Micro-photograph
PA
LDOCL
700 m
m
800 mm
18
Measured Small-Signal S-parameter
19
PA Performance vs VPA @60 GHz
5
10
15
20
25
30
0
3
6
9
12
15
0.7 0.8 0.9 1.0
Ps
at a
nd
P1d
B (
dB
m)
VPA (V)
PA
Em
ax (
%)
P1dB
Psat
PAEmax
20
0.01
0.1
1
10
100
1.E+001.E+021.E+041.E+061.E+081.E+10
Measured Lifetime of the PA ∆
I DS
at (%
)
Time (s)
VPA=1.00 V, Pout=10 dBm
VPA=1.00 V, Pout=5 dBm
VPA=0.75 V, Pout=5 dBm
1 102 104 106 108 1010
21
Measured Output Spectrum
-50
-40
-30
-20
-10
0
-2.5 -1.5 -0.5 0.5 1.5 2.5Centered Freq. (GHz)
Mag
nit
ud
e (
dB
)
-50
-40
-30
-20
-10
0
-2.5 -1.5 -0.5 0.5 1.5 2.5M
ag
nit
ud
e (
dB
) Centered Freq. (GHz)
(a)
IEEE 802.15.3c Spectrum mask
(b)
Spectrum centered at 62.64 GHz for QPSK modulation
(a) VPA=1.0 V, Pout=4 dBm; (b) VPA=0.7 V, Pout=3 dBm
22
Measured EVM for QPSK Modualtion
-20
-19
-18
-17
-16
-15
0.7 0.8 0.9 1
Pout = 5 dBm
Pout = 7 dBm
Pout = 10 dBm
VPA (V)
EV
M (
dB
)
23
60 GHz CMOS PA Performance Comparison
† Only for the last stage VPA
[6] M. Tanomura et. al, ISSCC 2008
[7] J. Chen et. al, ISSCC 2011
[5] A. Siligaris et. al, JSSC 2010
Ref. Process Vdd
(V)
P1dB
(dBm)
Psat
(dBm)
PAEmax
(%)
Lifetime
(year)
[5] 65 nm
SOI
1.2 7.1 10.5 22.3
N/A 1.8 12.7 14.5 25.7
2.6 15.2 16.5 18.2
[6] 90 nm 0.7 5.2 8.5 7.0 > 105*
1.0 10.5 11.5 8.5 > 10*
[7] 65 nm 1.0 15.0 18.6 15.1 N/A
[8] 65 nm 1.0 8.0 11.5 15.2 > 10*
This
work 65 nm
0.7† 5.8 10.1 8.1 > 102
1.0† 10.2 13.2 15.0 > 0.2
[8] W. L. Chan et. al, JSSC 2010
* Non-measured results
24
Conclusions
– The lifetime of the proposed PA can be improved
dramatically by dynamic operation.
– The tunable supply offers a possibility to meet
different linearity, efficiency, output power and
lifetime requirements in actual applications.
– The PA is insensitive to the process variations
thanks to the tunable supply voltage.
25
Thank you for your attention!
26
27
Output Power Distribution [5]
[5] A. Siligaris et. al, JSSC 2010
28
Transistor Measurement Data
[9] Q. Bu et. al, SSDM 2012
0
5
10
15
20
25
30
35
0 20 40 60 80 100
MA
G (
dB
)
Frequency (GHz)
CS
Cascode
29
Power Consumption
VPA IDigital IAnalog IPA
1.0 V 64 mA 312 mA 130 mA
0.7 V 64 mA 312 mA 120 mA
30
Lifetime Improvement of the PA
t1
t
t2 t3 tdtd-1
A1(t)1/n
A(t)1/n
A2(t)1/n
A3(t)1/n
Ad(t)1/n
Af(t)1/n
Fixed
Dynamic
Age function
31
The Flow Chart of Dynamic Operation
Sleep mode?
Bias on
Control frame
received?
Bias and output
power change
Yes
No
Yes
No
Bia
s a
nd
ou
tpu
t
po
we
r m
ain
tain
Bias off
32
Spectrum Measurement Setup
PA
AWG
PowerMeter
ParameterAnalyzer
ATT ATT
SpectrumAnalyzer
SignalGenerator
60-GHz TX Board
33
EVM Measurement Setup
PA 60-GHz RX Board
AWG Osci.
PowerMeter
ParameterAnalyzer
ATT ATT60-GHz TX Board
34
Dynamic Comparator Schematic [9]
Vi+ Vi-
Vb
ClkDi- Di+
Di- Di+
Vo+ Vo-
Vdd Vdd
[9] M. Miyahara et. al, ASSCC 2008
(a) First stage (b) Second stage
35
36
65 nm NMOSFET DC Stress Lifetime
1E+00
1E+01
1E+02
1E+03
1E+04
1E+05
1E+06
1E+07
1E+08
1E+09
1E+10
0.4 0.5 0.6 0.7 0.8 0.9
VGS=0.8 V
1/VDS (1/V)
VDS=1.2 V
VDS=2.0 V
100
101
102
103
104
105
106
107
108
109
1010 L
ife
tim
e t
(s
)
[2] E. Takeda et al., IEDL 1983
37
65 nm NMOSFET RF Stress Lifetime
∆I D
Sa
t (%
) VDS=1.2 V, VGS=0.8 V
1
10
100
1E+02 1E+03 1E+04 1E+05
Freq.=100 MHz,
Pout=11 dBm
102 103 104 105
Time (s) [3] L. Negre et al., JSSC 2012
38
CMOS PA Performance Comparison
† Only for the last stage VPA
[6] M. Tanomura et. al, ISSCC 2008
[7] J. Chen et. al, ISSCC 2011
[5] A. Siligaris et. al, JSSC 2010
Ref. Process Freq.
(GHz)
Vdd
(V)
P1dB
(dBm)
Psat
(dBm)
PAEmax
(%)
[5] 65 nm
SOI 60
1.2 7.1 10.5 22.3
1.8 12.7 14.5 25.7
2.6 15.2 16.5 18.2
[6] 90 nm 60 0.7 5.2 8.5 7.0
1.0 10.5 11.5 8.5
[7] 65 nm 60 1.0 15.0 18.6 15.1
[8] 65 nm 60 1.0 8.0 11.5 15.2
This
work 65 nm 60
0.7† 5.8 10.1 8.1
1.0† 10.2 13.2 15.0
[8] W. L. Chan et. al, JSSC 2010
39
CMOS PA Performance Comparison
† Only for the last stage VPA
[6] M. Tanomura et. al, ISSCC 2008
[7] J. Chen et. al, ISSCC 2011
[5] A. Siligaris et. al, JSSC 2010
Ref. Process Vdd
(V)
P1dB
(dBm)
Psat
(dBm)
PAEmax
(%)
Lifetime
(year)
[5] 65 nm
SOI
1.2 7.1 10.5 22.3
N/A 1.8 12.7 14.5 25.7
2.6 15.2 16.5 18.2
[6] 90 nm 0.7 5.2 8.5 7.0 > 105*
1.0 10.5 11.5 8.5 > 10*
[7] 65 nm 1.0 15.0 18.6 15.1 N/A
[8] 65 nm 1.0 8.0 11.5 15.2 > 10*
This
work 65 nm
0.7† 5.8 10.1 8.1 > 102
1.0† 10.2 13.2 15.0 > 0.2
[8] W. L. Chan et. al, JSSC 2010
* Estimation results