スライド 1 - Matsuzawa and Okada Laboratory

Measurement of Integrated PA-to-LNA
Isolation on Si CMOS Chip
Ryo Minami,JeeYoung Hong,
Kenichi Okada,and Akira Matsuzawa
Tokyo Institute of Technology, Japan
Matsuzawa
Matsuzawa
Lab.
& of
Okada
Lab.
Tokyo Institute
Technology
Outline
1
• Tx leakage and problem
• Evaluation of Tx leakage
– Tx leakage paths
– Measurement and Simulation method
• Result
• Conclusion
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Background
2
Conventionally, Power Amplifier (PA) has been
implemented by compound semiconductors.
• Merit
– High frequency characteristic
– High supply voltage
• Demerit
– Chip area and cost
CMOS technology has been developed.
CMOS transistors can provide a sufficient performance for
PA design.
The level of Tx leakage increases.
To investigate Tx leakage paths.
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Tx leakage and problem
3
LNA nonlinearity
FDD system
Tx
leakage
interference
signal
LNA
Tx
leakage
Rx
signal
received
signal
Tx
leakage
Duplexer
transmitted
signal
PA
cross
modulation
interference
signal
Rx
signal
intermodulation
distortion
• The large transmitted signal leaks to Rx input side.
• Tx leakage causes IM and degrades demodulation quality.
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Tx leakage paths
4
The integration of the PA with other blocks
increases many coupling paths.
conventional
single-chip
LNA
LNA
Duplexer
Duplexer
PA
2010/12/10
substrate coupling
PA
inductor coupling
wire coupling (Vdd, GND)
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Dicing
5
Magnetic
PA
LNA
gnd
VDD
gnd
VDD
-+
p+
n+
-+
n+
p+
n+
p+
n+
n+
Substrate coupling and
inductor coupling are
generated because PA
and LNA are integrated
on the same chip.
n+
Substrate coupling
Magnetic
PA
LNA
gnd
VDD
gnd
VDD
-+
p+
n+
-+
n+
p+
n+
n+
p+
n+
n+
Substrate coupling is
blocked by physically
separating the PA and
LNA[1].
[1] J.Y. Hong et al., IEICE society Conference, 2009.
2010/12/10
Air gap
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Measurement method
6
-5dBm
-5dBm
Network Analyzer
Network Analyzer
• substrate coupling
• inductor coupling
• inductor coupling
probe
air gap
probe
PA
LNA1
LNA2
probe
LNA3
probe
LNA4
PA
Chip
LNA2
LNA1
LNA3
LNA4
Chip
absorber
absorber
isolation
isolation[dB]  S 21  (GainPA  GainLNA )
couplingsubstrate  dicingbefore  dicingafter
PA
On-Chip
LNA
output
input
GLNA
GPA
S21
After dicing, the distance between PA and LNA
is 0.74mm, 1.55mm, 2.37mm, 3.20mm.
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Specification of the designed PA and LNA 7
LNA
PA
PA
LNA
Technology
0.18 mm CMOS process
Frequency
5 GHz
VDD
3.3 V
1.8 V
Gain at 5 GHz
5.5 dB
15.1 dB
NF at 5 GHz
Area
2010/12/10
2.7 dB
1.01mm×1.01mm 0.70mm×1.01mm
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Simulation method
8
Magnetic coupling simulation by HFSS (an electromagnetic
field simulator)
Circuit simulation with magnetic coupling effect by goldengate
LNA
inductor
1.8 V
PA
inductor
3.3 V
Magnetic coupling
(2.5 turn)
(1.5 turn)
output
distance
input
On-chip
50 W
104mm
PA
LNA
0.74mm
1.55mm
2.37mm
3.20mm
160mm
Assume
The magnetic coupling between the inductor at PA output
side and the inductor at LNA input side is the most dominant.
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Simulation result
9
-40
Gain of PA
and LNA
-60
S21[dB]
-80
-100
-120
0.74mm
2.37mm
-140
1.55mm
3.2mm
-160
0
1
2
3
4
5
6
7
8
9
10
Frequency[GHz]
• Gain of PA and LNA are shown at 5 GHz
• S21 depends on the distance between PA
and LNA
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Measurement result
S21 [dB]
noise floor
or
thermal noise
10
-40
sim.: dotted
-60
probe
coupling
meas.: solid
-80
-100
-120
0.74mm
2.37mm
-140
1.55mm
3.2mm
-160
0
Assume
1
2
3
4
5
6
frequency[GHz]
7
8
• Under 3 GHz
– Thermal noise from the resistance of LNA
– Noise floor of the network analyzer
• Around 5 GHz
– Noise floor of the network analyzer with gain
• Over 8 GHz
– Probe coupling exists.
2010/12/10
R. Minami , Tokyo tech.
9
10
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Noise floor and probe coupling
● Probe coupling
● Noise floor
-5dBm
Network
Analyzer
-5dBm
3cm
probe
11
probe
open(=S21)
Network
Analyzer
0.98mm,1.8mm,2.62mm,3.45mm
probe
2cm
probe
coupling(=S21)
0.3mm
absorber
absorber
Noise floor
S21 with 3 cm distance
between probes and 2cm
distance between probe and
absorber
Probe coupling
S21 with 4 patterns of
probe distance and
0.3mm distance between
probe and absorber
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Probe coupling problem
12
● The case of measuring chip
coupling
signal
probe
PA
probe
LNA
signal
Chip
GND
Signal flows into
the chip.
absorber
● The case of measuring probe coupling
probe
coupling
signal
probe
0.3mm
absorber
2010/12/10
R. Minami , Tokyo tech.
This is the worst
case of probe
coupling.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Thermal noise from the resistance of LNA 13
-5dBm
Network
Analyzer
(a): noise calculation point
×
On-Chip
LNA
PA
input
output
GPA
k: 1.38*10-23 [J/K]
T: 300 [K]
GLNA
B: 10 [Hz]
Cable loss:2 [dB]
Noisethermal [ dB]  10Log(kTB)  NFLNA  GainLNA  Losscable  NAout
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Result of error analysis (higher and lower side) 14
S21&Thermal noise [dB]
-40
0.74mm
probe couping
noise floor NA
noise thermal
-60
-80
-100
-120
-140
-160
0
1
2
3
7
6
5
4
frequency[GHz]
• Under 3 GHz
– Noise floor of the network analyzer
• Over 8 GHz
– Probe coupling exists.
2010/12/10
R. Minami , Tokyo tech.
8
9
10
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
Result of error analysis (middle section)
-40
sim.: dotted
meas.: solid
-60
GPA+GLNA
S21 [dB]
-80
-100
-120
0.74mm
1.55mm
2.37mm
3.2mm
noise floor
-140
-160
0
1
2
3
4
5
6
7
frequency[GHz]
8
9
10
• Around 5 GHz
– Noise floor of network analyzer with gain
noisefloor(5GHz)  GainPA  GainLNA  80dB
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
15
Comparison of coupling data at 5 GHz
16
-40
Isolation [dB]
-50
• The isolation value of
inductor coupling is
affected by the noise
floor of the NA over
1.8mm distance
between PA and LNA.
inductor coupling
-60
simulation
-70
-80
-90
-100
noise floor
-110
-120
1
2
distance[mm]
• Total isolation value
becomes smaller than
that of the duplexer over
0.4mm
• Substrate coupling is
the most dominant factor
in Tx leakage.
2010/12/10
3
4
-40
Duplexer isolation
-50
Isolation [dB]
0
-60
all Tx leakage
-70
inductor coupling
substrate coupling
-80
-90
-100
-110
-120
0
R. Minami , Tokyo tech.
1
2
3
distance[mm]
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
4
Conclusion
17
• Substrate coupling is the most dominant
factor in Tx leakage for Si substrate.
• Total isolation value becomes smaller than
that of the duplexer over 0.4mm distance
between PA and LNA.
• Error sources in measurement are noise
floor of the network analyzer and probe
coupling.
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology
18
Thank you for your attention!
2010/12/10
R. Minami , Tokyo tech.
Matsuzawa
Matsuzawa
Lab.
& Okada
Lab.
Tokyo Institute
of Technology