modeling and simulation of jitter in phase-locked …...modeling and simulation of jitter in...

28
Modeling and Simulation of Jitter in Phase-Locked Loops due to Substrate Noise Jae Wook Kim, Yi-Chang Lu and Robert W. Dutton Stanford University September 22, 2005

Upload: others

Post on 19-Feb-2020

13 views

Category:

Documents


0 download

TRANSCRIPT

Modeling and Simulation of Jitter in Phase-Locked Loops due to

Substrate Noise

Jae Wook Kim, Yi-Chang Lu and Robert W. DuttonStanford University

September 22, 2005

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

Substrate Noise in Systems-on-a-Chip (SoC)

Demands for total system on same chipMixed signal systems: substrate noise issueLNA, VCO, PA and DA/AD as victims

Substrate (Si)

Digital Circuits

Substrate Noise

Analog Circuits

Simulations for Analog Systems under Substrate Noise Environment

The simulations of substrate noise coupling to analog systems can be simple conceptually, but is definitely demanding in transistor-level and transient analysis mode.The blocks of an analog system are not equally sensitive to substrate noise. And substrate noise coupling mechanism itself is deterministic, thus allowing to abstract behavioral models.Considering these factors, simulations for PLL under substrate noise environment can be a challenging but good candidate.

Substrate Noise Coupling in SoC- Block diagram of mixed signal system with PLL

Current Noise

Noise Couplingto VCO

1.85 1.9 1.95 2x 10-8

0

5

10

15

x 10-5

TIME [sec]

Cur

rent

[A]

Ivdd of an inverter

Mathematical modelHspice simulation

3.35 3.4 3.45 3.5x 10-8

0

1

2

TIME [sec]

Cur

rent

[A]

Ignd of an inverterMathematical modelHspice simulation

1.85 1.9 1.95 2 2.05x 10-8

-1

0

1

2

3

4

5x 10-3

TIME [sec]

Vol

tage

[V]

VdpbulkHspice simulationMathematical model

3.35 3.4 3.45 3.5 3.55x 10-8

-8

-6

-4

-2

0

2x 10-3

TIME [sec]

Volta

ge [V

]

VdnbulkHspice simulationMathematical model

0 0.5 1 1.5 2 2.5x 10

-5

0.6

0.7

TIME [sec]

Vol

tage

[V]

Control Voltage of VCO

4 6 8 10 12x 107

-60

-40

-20

0

20

40

Frequency [Hz]Pow

er S

pect

ral D

ensi

ty [d

Bc/

HPower Spectrum of PLL with DNE noise

0 0.2 0.4 0.6 0.8 1x 10-7

-10

-5

0

5 x 10-3

TIME [sec]

Vol

tage

[V]

themetical model of substrate noise at an

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.60

0.5

1

1.5

2

2.5

3 x 108

Voltage [V]

Freq

uenc

y [H

z]

Charcteristic curve between a control voltage and a frequency of VCO

Hspice simulationCurve-fitted polynomial

10 15 20 25 30 35-4

-3

-2

-1

0

1

2

3 x 10-12

Time when the impulse applied [nsec]

Pha

se s

hift

[sec

]

ISF15th rising edge15th falling edge

τττφ dvthtt

substrate )(),()( ∫∞−

=∆

Control Voltage

ISF

Output Spectrum

Substrate Noise

Voltage Noise

Coupling throughSubstrate

Two Levels of Simulation

Digital Noise Emulator and PLLTargetSystem

C++, Verilog-AHspice, SpectreTools

Quick guideFast but rough

VerificationSlow but accurate

Purpose / Features

System levelCircuit level

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

Substrate Noise Coupled to VCO

gnd

in out

bias

Vdd

nbulk

Converter

+

-

Inverter

1.4595 1.46 1.4605 1.461 1.4615 1.462 1.4625 1.463 1.4635 1.464

x 10-8

0

0.02

0.04

0.06

0.08

0.1

TIME [sec]

Nod

e V

olta

ge [V

]

Voltage Impulse at nbulk node

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

x 10-7

-0.5

0

0.5

1

1.5

2

TIME [sec]

Out

put V

olta

ge [V

]

Output Voltage at out node

1.552 1.5522 1.5524 1.5526 1.5528 1.553 1.5532 1.5534 1.5536 1.5538

x 10-7

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

TIME [sec]

Out

put V

olta

ge [V

]

Output Voltage at out node (Enlarged)w\o impulsew\ impulse

Substrate Noise Sensitivity Function of VCO

Converter

DC

PULSE

+

-

9.0n 10.0n 11.0n 12.0n 13.0n

-40.0m

-20.0m

0.0

20.0m

40.0m

60.0m

80.0m

100.0m

120.0m

140.0m

Volta

ge [V

]

Time when the impulse applied [sec]9.0n 10.0n 11.0n 12.0n 13.0n

-100.0f

-50.0f

0.0

50.0f

100.0f

150.0f

200.0f

Phas

e Sh

ift [s

ec]

Time when the impulse applied [sec]

-300.0m-200.0m-100.0m 0.0 100.0m 200.0m 300.0m

-6.0p

-4.0p

-2.0p

0.0

2.0p

4.0p

6.0p

Phas

e Sh

ift [s

ec]

Applied Voltage Amplitude [V]

15th Rising Edge 15th Falling Edge

Behavioral Model – Noise coupling equation

• ISFV – Impulse Sensitivity Function between substrate noise and phase shift

10.0n 15.0n 20.0n 25.0n 30.0n 35.0n

-2.5p

-2.0p

-1.5p

-1.0p

-500.0f

0.0

500.0f

Phas

e Sh

ift [s

ec]

Time when the impulse applied [sec]∫=t

dISFV∆Φ(t)0 vsub )()( τττ

Converter

DC

PULSE

+

-

1 1.05 1.1 1.15 1.2 1.25 1.3 1.35x 10-8

-4

-3

-2

-1

0

1

2

3

4 x 10-4

Time [sec]

Cur

rent

pul

led

out [

A]

Current Disturbance due to Pulse Noise

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

x 10-7

-0.5

0

0.5

1

1.5

2

TIME [sec]

Out

put V

olta

ge [V

]

Output Voltage at out node

1.4595 1.46 1.4605 1.461 1.4615 1.462 1.4625 1.463 1.4635 1.464

x 10-8

0

0.02

0.04

0.06

0.08

0.1

TIME [sec]

Nod

e V

olta

ge [V

]

Voltage Impulse at nbulk node

Analysis - Substrate noise coupling mechanism to VCO

Converter

DC

PULSE

+

-

Substrate Noise toCurrent Disturbance

Current Disturbanceto Phase Shift

1.552 1.5522 1.5524 1.5526 1.5528 1.553 1.5532 1.5534 1.5536 1.5538

x 10-7

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

TIME [sec]

Out

put V

olta

ge [V

]

Output Voltage at out node (Enlarged)w\o impulsew\ impulse

Analysis - Substrate noise to current disturbance

Body effectLinear parameter: gmb

∆I = gmbVsub

Capacitive couplingThrough pn+ junction capacitance∆I = CjdVsub/dt

Total current noise∆I = gmbVsub+CjdVsub/dt

Bulkn+ n+

SourceDrain

Gate

CapacitiveCoupling

BodyEffect - gmb

Substrate Noise - Vsub

Analysis- Current disturbance to phase noise: ISFi

10.0n 15.0n 20.0n 25.0n 30.0n 35.0n

-1.5p

-1.0p

-500.0f

0.0

500.0f

1.0p

1.5p

2.0p

Phas

e Sh

ift [s

ec]

Time when the impulse applied [sec]

∫+=

=

t

t

dτISFdτdVCISFgV

dτ(τISF∆I(τ)∆Φ(t)

0 isub

jimbsub

0 i )

0)(

),(function impulsean isWhen

00

0ij0sub

j

0sub

== ∫∫tt

i dd

)-tδ(dtISFCdISFddVC

)δ(t-tV

ττττ

τ

10.0n 15.0n 20.0n 25.0n 30.0n 35.0n

-2.5p

-2.0p

-1.5p

-1.0p

-500.0f

0.0

500.0f

1.0p

1.5p

2.0p

Phas

e Sh

ift [s

ec]

Time when the impulse applied [sec]

ISFi

ISFv

Behavioral Model – Noise coupling equation based on analysis

∫∫

∫∫

+=

+=+=

tt

tt

i

dτISFdτdVCdτISFV

dτISFdτdVCISFg(Vdτ)ISF

dτdVCg(V∆Φ(t)

0 isub

j0 vsub

0 isub

jimbsub0sub

jmbsub )

10.0n 15.0n 20.0n 25.0n 30.0n 35.0n

-2.5p

-2.0p

-1.5p

-1.0p

-500.0f

0.0

500.0f

Phas

e Sh

ift [s

ec]

Time when the impulse applied [sec]

ISFix gmb ISFv

imbv ISFgISF =

∫=t

dISFV∆Φ(t)0 vsub )()( τττ

∫=t

dτ(τISFgV∆Φ(t)0 imbsub )

Behavioral Model – Block diagram of VCO

1 1.5 2 2.5 3 3.5x 10-8

-2.5

-2

-1.5

-1

-0.5

0

0.5x 10-12

Time when a voltage pulse applied [sec]

Phas

e S

hift

[sec

]

ISF

ISFv

integration

τdt

∫∞−

Noisy Phase to Divider

Substrate Noise

0 0.2 0.4 0.6 0.8 1x 10-7

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

TIME [sec]

Vol

tage

[V]

Substrate Noise at nbulk

VCO

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.60

0.5

1

1.5

2

2.5

3 x 108

Voltage [V]

Freq

uenc

y [H

z]

Charcteristic curve between a control voltage and a frequency of VCO

Hspice simulationCurve-fitted polynomial

Ideal Phase

2 2.1 2.2 2.3 2.4x 10-7

0

1

2

3

4

5

6

TIME [sec]

Pha

se[R

ad]

Φ(t)

derivative

dtdCjnction

1 1.5 2 2.5 3 3.5x 10-8

-1.5

-1

-0.5

0

0.5

1

1.5

2x 10-12

Time when a current pulse applied [sec]

Phas

e S

hift

[sec

]

ISFi

ISFi

Phase shift

0 0.5 1 1.5 2 2.5x 10-5

-4.5-4

-3.5-3

-2.5-2

-1.5-1

-0.50

0.5

TIME [sec]

Pha

se S

hift

[Rad

]

∆Φ(t)

0 0.5 1 1.5 2 2.5x 10

-5

0.6

0.7

TIME [sec]

Vol

tage

[V]

Control Voltage of VCO

Control Voltage

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

Comparison between Simulation Methodologies - I

Verilog-A and HspiceTwo versions of Verilog-A

Interpreter based: CadenceCompiler based: Tiburon Design Automation

Transient simulation of PLL and sampling 1,250,000 temporal points

18 mins19 mins14 hrs 53 mins

Simulation time

Tiburon Verilog-A

Cadence Verilog-AHspiceLanguage

Comparison between Simulation Methodologies - II

81MHz PLL with white Gaussian random noise

Summation of three sinusoidal waves (5MHz, 13MHz, and 90MHz )

70M 80M 90M 100M

-60.0

-40.0

-20.0

0.0

20.0

40.0

Pow

er S

pect

ral D

ensi

ty [d

B/H

z]

Frequency [Hz]

Hspice Verilog-A

70M 80M 90M 100M

-60.0

-40.0

-20.0

0.0

20.0

40.0

Pow

er S

pect

ral D

ensi

ty [d

B/H

z]

Frequency [Hz]

Hspice Verilog-A

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

Period Histogram- Concept

A metric to estimate the phase shiftIdeal case

One period, one peak With noise

Jitter causes the histogram changing shape.

Ideal case

With noise2 2.2 2.4 2.6 2.8 3

x 10-9

0

1000

2000

3000

4000

5000

6000

period [sec]

HIS

T

2 2.2 2.4 2.6 2.8 3x 10-9

0

50

100

150

Period

hist

Period Histogram- Noise dependency

Periodic noise400MHz PLL with 100mV, 100MHz sinusoidal wave noisePeak splitting

Random noiseWith 100mV(RMS) white Gaussian noiseHistogram spreading

2.3 2.35 2.4 2.45 2.5 2.55 2.6 2.65 2.7x 10-9

0

50

100

150

Period

hist

2 2.2 2.4 2.6 2.8 3x 10-9

0

200

400

600

800

1000

1200

1400

period [sec]

HIS

T

0 0.2 0.4 0.6 0.8 1x 10-8

-4

-2

0

2x 10-12

0 0.2 0.4 0.6 0.8 1x 10-8

-0.1

0

0.1

0 0.2 0.4 0.6 0.8 1x 10-8

-1

0

1

x 10-10

ISF ( of three NMOS devices)

100MHz Sine Wave Noise

∆Φ = dtftISF

MHzofpeiord∫ ×

4001

)2sin(1.0 π

1/400MHz

Period Histogram- Variation with phase of noise

400MHz PLL with 100MHz sinusoidal noise

2 2.2 2.4 2.6 2.8 3x 10-9

0

200

400

600

800

1000

1200

1400

period [sec]

HIS

T0 0.2 0.4 0.6 0.8 1

x 10-8

-4

-2

0

2x 10-12

0 0.2 0.4 0.6 0.8 1x 10-8

-0.1

0

0.1

0 0.2 0.4 0.6 0.8 1x 10-8

-1

0

1

x 10-10

ISF ( of three NMOS devices)

100MHz Sine Wave Noise

∆Φ = dtftISF

MHzofpeiord∫ ×

4001

)2sin(1.0 π

1/400MHz

2 2.2 2.4 2.6 2.8 3x 10-9

0

500

1000

1500

2000

2500

period [sec]

HIS

T

2.4

2.45

2.5

2.55

2.6

x 10-9

0

50

100

150

200

250

300

350

020004000

Period [s]Phase [deg]

HIS

T

Period Histogram- 3-D plot of period histogram

Switching noise100MHz sinusoidal noise

22.2

2.4

2.62.8

3

x 10-9

0

100

200

300

0

2000

4000

Period [sec]Phase [deg]

HIS

T 6.8 7 7.2 7.4 7.6 7.8x 10

-8

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

Time [s]

Subs

trate

Noi

se [V

]

5ns

bestworst

Test Vehicle200MHz PLL* with Digital Noise Emulator (DNE)

TSMC .18µm generic logic process (TSMC CL018G)Period of the PLL output is measured and plotted as 3-D period histogram.

*Designed by Barcelona Design

DNE

PLL

Measured 3-D plots of period histogram

Comparison with Measurements- 3-D Plot of period histogram

Measurement with 160MHz clock noise

Verilog-A Simulation with 160MHz clock noise plus random noise

Phase [deg]

Period [sec]

HIS

T

Phase [deg]Period [sec]

HIS

T

Outline

Introduction

Substrate Noise Coupling Mechanism and Modeling Technique

Comparison between Simulation Methodologies

Period Histogram and Comparison with Measurements

Conclusion

ConclusionImpulse Sensitivity Function (ISF) can be used as the behavioral model of the substrate noise coupling to VCO with a linear and time variant approach.Two versions of Verilog-A are utilized to implement the system level simulation and compared with the transistor level simulation (Hspice).Results from the system level simulation are compared with the measurement results in terms of the period histogram, showing good correlation.