RFIC - Long Beach June 12-14, 2005 1
(RTU4A-1)
The Effect of Substrate Noise on VCO Performance
Nisha Checka, David D. Wentzloff, AnanthaChandrakasan, Rafael Reif
Microsystems Technology Laboratory, MIT60 Vassar St. Rm. 39-625
Cambridge, MA, (USA) 02139
RFIC - Long Beach June 12-14, 2005 2
Outline
• Motivation/Background
• Noise Characterization Test Chip
• VCO Results– Effect of Bias Current– Effect of Guard Rings
• Conclusions
RFIC - Long Beach June 12-14, 2005 3
What is Substrate Noise?• Digital circuits inject noise into substrate during
switching
• Noise propagates to analog section via substrate• Effect on analog circuits: substrate contacts, pickup
through large capacitive nodes, backgate effect
RFIC - Long Beach June 12-14, 2005 4
Motivation
Digital OFF Digital ON
• UWB transceiver chip– TSMC 0.18 µm mixed mode process
RFIC - Long Beach June 12-14, 2005 5
• TSMC 0.18 µm mixed-mode process (non-epi substrate)
• ρsub ≈ 10-15 Ωcm• Triple-well process (~ 20 dB isolation)
900 MHZ GR2.4
GHZGR2
2.4GHZGR1
2.4GHZ
5.2 GHZGR
5.2
GH
Z900 MHz
900M n+
2.4G n+5.2G n+
p+ sense
Die Microphotograph of Test Chip
Noise Characterization Test Chip
RFIC - Long Beach June 12-14, 2005 6
• Multiple injection/sensing locations• Three VCO center frequencies
Experimental Setup
RFIC - Long Beach June 12-14, 2005 7
• VCO center frequency– 900 MHz, 2.4 GHz, 5.2 GHz
• VCO bias current
• Isolation– No guard ring, guard ring
Parameters
RFIC - Long Beach June 12-14, 2005 8
Noise Coupling Paths
substrate
substrate noise
inductor-substrate
capacitance
Vout
VDD/GNDIVCO
VCTRL
RFIC - Long Beach June 12-14, 2005 9
Noise couples through GND line and inductor-substrate capacitance
5.2 GHz VCO (no GR)
-105
-85
-65
-45
-25
4.306 4.310 4.314 4.318
Frequency (GHz)
Pow
er (d
Bm
)
VCO offVCO on
P=-60 dBm
P=-72.2 dBm
12.2 dB gain by powering up
Noise injection frequency
Noise Reception
sub sub
Vcont
IVCO
RFIC - Long Beach June 12-14, 2005 10
900 MHz2.4 GHz
5.2 GHz
Noise coupling through inductor decreases
Inductors
L = 11.4 nHC = 220 fFA ≈ 340 x 340 µm2
L = 3.19 nHC = 100 fFA ≈ 220 x 220 µm2
L = 655 pHC = 38 fFA ≈ 140 x 140 µm2
RFIC - Long Beach June 12-14, 2005 11
Inductor Noise Coupling
-105
-95
-85
-75
-65
-55
-45
899.6 899.7 899.8 899.9 900 900.1 900.2 900.3 900.4
Frequency (MHz)
Pow
er (d
Bm
)
5.2 GHz VCO900 MHz VCO
Noise -- Inductor Component
13.2 dB
RFIC - Long Beach June 12-14, 2005 12
• VCO center frequency– 900 MHz, 2.4 GHz, 5.2 GHz
• VCO bias current
• Isolation– No guard ring, guard ring
Parameters
RFIC - Long Beach June 12-14, 2005 13
ILOW = 1.81 mAIMID = 2.71 mAIHIGH = 3.41 mA
IBUFF = 2.73 mA
Ratio of carrier to noise power for 5.2 GHz VCO
Effect of VCO Bias Current5.2 GHz VCO Phase Noise
-85
-75
-65
-55
-45
-35
-25
-15
-5
Frequency Offset
Leve
l (dB
c/H
z)
LowMidHigh
100 kHz 1 MHz 10 MHz
23.1 dB3.41 mA26.3 dB2.71 mA9.9 dB1.81 mA
Pc/Pn (dB)IVCO
RFIC - Long Beach June 12-14, 2005 14
• VCO center frequency– 900 MHz, 2.4 GHz, 5.2 GHz
• VCO bias current
• Isolation– No guard ring, guard ring
Parameters
Dual guard rings
RFIC - Long Beach June 12-14, 2005 15
Guard rings
Effect of Guard Rings
• Guard rings only surround active devices− Can only attenuate ground component of noise
RFIC - Long Beach June 12-14, 2005 16
Guard Ring
-100
-95
-90
-85
-80
-75
-70
-65
-60
899 899.5 900 900.5 901Frequency (MHz)
Pow
er (d
Bm
)
-71.5 dBm
No Guard Ring
-100
-95
-90
-85
-80
-75
-70
-65
-60
899 899.5 900 900.5 901Frequency (MHz)
Pow
er (d
Bm
)
-71 dBm
• Guard ring has no effect on inductor component
Test: VCO powered off. Noise injected at 900 MHz
Guard Rings and Inductor-Sub Noise
RFIC - Long Beach June 12-14, 2005 17
Effect of Guard Ring for 5.2 GHz VCO
Received Noise for 5.2 GHz VCO
-35
-30
-25
-20
-15
-10
-5
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2foffset (MHz)
Pow
er (d
Bc)
no GRGR
RFIC - Long Beach June 12-14, 2005 18
Guard Ring Attenuation
-25
-20
-15
-10
-5
0
-1.5 -0.5 0.5 1.5 2.5 3.5 4.5
foffset (MHz)
Pow
er (d
Bc)
900M900M IM15.2G IM15.2G
• 900 MHz: -23.4 dB to -13.1 dB
• 5.2 GHz: -13.5 dB to 0.64 dB
Guard Ring Attenuation
5.2 GHz IM15.2 GHz
900 MHz IM1
900 MHz
RFIC - Long Beach June 12-14, 2005 19
5.2 GHz VCO Spectrum
-105
-85
-65
-45
-25
4.3145 4.3149 4.3153 4.3157 4.3161 4.3165
Frequency (GHz)
Pow
er (d
Bm
)
No NoiseFree-running VCO
• As fnoise approaches fVCO, and if Pnoise is comparable to Pcarrier, VCO can lock to fnoise
VCO Locking
RFIC - Long Beach June 12-14, 2005 20
VCO Locking
• As fnoise approaches fVCO, and if Pnoise is comparable to Pcarrier, VCO can lock to fnoise
5.2 GHz VCO Spectrum
-105
-85
-65
-45
-25
4.3145 4.3149 4.3153 4.3157 4.3161 4.3165
Frequency (GHz)
Pow
er (d
Bm
)
LockedVCO locked tofnoise = 4.3153 GHz
RFIC - Long Beach June 12-14, 2005 21
• As fnoise approaches fVCO, and if Pnoise is comparable to Pcarrier, VCO can lock to fnoise
5.2 GHz VCO Spectrum
-105
-85
-65
-45
-25
4.3145 4.3149 4.3153 4.3157 4.3161 4.3165
Frequency (GHz)
Pow
er (d
Bm
)
No NoiseLocked
Free-running VCOVCO locked tofnoise = 4.3153 GHz
Locked to noise 20 kHz away
VCO Locking
RFIC - Long Beach June 12-14, 2005 22
Resonant Gain Behavior
Received Noise for 5.2 GHz VCO
-35
-30
-25
-20
-15
-10
-5
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2foffset (MHz)
Pow
er (d
Bc)
no GRGR
RFIC - Long Beach June 12-14, 2005 23
20 kHz4.3137-4.3139 GHz4.314 GHzGR
50 kHz4.3145-4.315 GHz4.3148 GHzNo GR
RangeLocking FrequenciesfcenterNo GR/GR
5.2 GHz VCO and Injection Locking
30 kHz2.03065-2.03068 GHz2.031 GHzGR
50 kHz2.0275-2.02755 GHz2.027 GHzNo GR
RangeLocking FrequenciesfcenterNo GR/GR
2.4 GHz VCO and Injection Locking
-Doesn’t Lock805.13 MHzGR
5 kHz812.9-812.905 MHz812.89 MHzNo GR
RangeLocking FrequenciesfcenterNo GR/GR
900 MHz VCO and Injection Locking
Guard rings improve VCO locking
VCO Locking Range
carrier
noiseolock I
IQff2
≈
[1] (Razavi, JSSC, 39(9))
[1]
RFIC - Long Beach June 12-14, 2005 24
Conclusions
• Phase noise of a VCO is adversely affected by substrate noise– In extreme, VCO can lock to noise
• Bias current plays important role
• Guard rings are effective at lower frequencies, less useful at higher frequencies