HSPICE® User Guide:

Advanced Analog
Simulation and Analysis
Version N-2017.12, December 2017
Copyright and Proprietary Information Notice
© 2017 Synopsys, Inc. This Synopsys software and all associated documentation are proprietary to Synopsys, Inc. and may only be
used pursuant to the terms and conditions of a written license agreement with Synopsys, Inc. All other use, reproduction,
modification, or distribution of the Synopsys software or the associated documentation is strictly prohibited.
Destination Control Statement
All technical data contained in this publication is subject to the export control laws of the United States of America.
Disclosure to nationals of other countries contrary to United States law is prohibited. It is the reader’s responsibility to
determine the applicable regulations and to comply with them.
Disclaimer
SYNOPSYS, INC., AND ITS LICENSORS MAKE NO WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, WITH
REGARD TO THIS MATERIAL, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
Trademarks
Synopsys and certain Synopsys product names are trademarks of Synopsys, as set forth at
https://www.synopsys.com/company/legal/trademarks-brands.html.
All other product or company names may be trademarks of their respective owners.
Third-Party Links
Any links to third-party websites included in this document are for your convenience only. Synopsys does not endorse
and is not responsible for such websites and their practices, including privacy practices, availability, and content.

Synopsys, Inc.
690 E. Middlefield Road
Mountain View, CA 94043
www.synopsys.com

ii HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Contents

Related Products and Trademarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

Topics in This Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
The HSPICE Documentation Set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi
Customer Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii

1. HSPICE Advanced Analog Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Introduction to HSPICE Advanced Analog Analyses . . . . . . . . . . . . . . . . . . . . 1
HSPICE Advanced Analog Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
HSPICE Advanced Analog Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Recommended Analysis Method for Typical RF Design Blocks. . . . . . . . 5
HSPICE Advanced Analog Output Files . . . . . . . . . . . . . . . . . . . . . . . . . 6
Using Custom WaveView . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Creating a Configuration File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Inserting Comments in a .hspicerf File. . . . . . . . . . . . . . . . . . . . . . . 11
Using Wildcards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Limiting Output Data Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Generating Measurement Output Files . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Advanced Analog Demonstration Input Files . . . . . . . . . . . . . . . . . . . . . . 13
Use of Example Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Part I: Periodic Steady-State Analysis

2. Harmonic Balance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Tutorial Examples Harmonic Balance Analysis . . . . . . . . . . . . . . . . . . . . . . . . 17
Example 1: Using HB Analysis for a Power Amplifier. . . . . . . . . . . . . . . . 18
Example 2: Using HB Analysis for a Low Noise Amplifier . . . . . . . . . . . . 22
Harmonic Balance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Features Supported . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Harmonic Balance Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

iii
Contents

Steady-State HB Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Harmonic Balance Output Measurements . . . . . . . . . . . . . . . . . . . . . . . . 32
HB .PRINT and .PROBE Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . 34
HB Output Data Files and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
HB Error and Warning Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
HB Analysis Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
HB Analysis Warning Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Calculating Power Measurements After HB Analyses . . . . . . . . . . . . . . . 39
Power Dissipated in a Resistor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Power Delivered to a Port Element. . . . . . . . . . . . . . . . . . . . . . . . . . 40
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Calculations for Time-Domain Output . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Using .MEASURE with .HB Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Troubleshooting Convergence Issues with Harmonic Balance Analysis . . . . . 46
Phase Differences Between HB and SIN Sources . . . . . . . . . . . . . . . . . . . . . 47
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Optimizing with HB Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3. Shooting Newton Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Shooting Newton Steady-State Time Domain Analysis (.SN) . . . . . . . . . . . . . 51
Shooting Newton Analysis — Tutorial Example . . . . . . . . . . . . . . . . . . . . 52
Shooting Newton Analysis — Setup. . . . . . . . . . . . . . . . . . . . . . . . . 53
Clock Buffer Analysis Example. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
SN .PRINT and .PROBE Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . 59
Shooting Newton with Fourier Transform (.SNFT) . . . . . . . . . . . . . . . . . . . . . . 61
Other Shooting Newton Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
SN Error and Warning Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Troubleshooting Convergence Issues with Shooting Newton Analysis . . . . . . 63

iv
 Contents

Part II: Oscillator and PLL Analysis

4. Oscillator Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Oscillator Analysis (.HBOSC and .SNOSC). . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Tutorial Examples Using HBOSC Analysis . . . . . . . . . . . . . . . . . . . . . . . 69
Example 1 — Colpitts Oscillator. . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Example 2 — Using HBOSC for a CMOS GPS VCO. . . . . . . . . . . . 75
.HBOSC Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Using the .MEASURE Command with .HBOSC . . . . . . . . . . . . . . . . . . . 83
HB Simulation of Ring Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Troubleshooting Convergence Problems . . . . . . . . . . . . . . . . . . . . . . . . . 84
General Convergence Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Outer Loop Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Inner Loop Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Shooting Newton Oscillator Analysis (.SNOSC) . . . . . . . . . . . . . . . . . . . . . . . 90
Ring Oscillator Example. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5. Phase Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

Introduction to Phase Noise Analysis (.PHASENOISE). . . . . . . . . . . . . . . . . . 97
Identifying Phase Noise Spurious Signals . . . . . . . . . . . . . . . . . . . . . . . . 100
Phase Noise Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
PHASENOISE Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Whole Circuit and Specified Element Phase Noise . . . . . . . . . . . . . 103
Frequency-Dependent and Frequency-independent Sources . . . . . 104
Frequency and Bias Dependencies. . . . . . . . . . . . . . . . . . . . . . . . . 105
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Measuring Phase Noise with .MEASURE PHASENOISE . . . . . . . . . . . . 106
Optimizing with PHASENOISE Measurement . . . . . . . . . . . . . . . . . . . . . 108
Amplitude Modulation/Phase Modulation Separation . . . . . . . . . . . . . . . 108
AM/PM .PRINT and .PROBE Statement Syntax . . . . . . . . . . . . . . . 109
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Output Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Important Note for AM/PM Users . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Interpreting Phase Noise Analysis Results . . . . . . . . . . . . . . . . . . . . . . . 111
Accumulated Jitter Measurement for Closed Loop PLL Analysis . . . . . . . . . . 113
Jitter Measurements from Phase Noise . . . . . . . . . . . . . . . . . . . . . . . . . . 113

v
Contents

Jitter Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

Jitter Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
.MEASURE Statements for Jitter . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Peak-to-Peak Jitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Errors/Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Clock Source with Random Jitter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

6. Behavioral PLL Noise Analysis (.ACPHASENOISE) . . . . . . . . . . . . . . . . . . 121

ACPHASENOISE Analysis .PRINT and .PROBE Syntax . . . . . . . . . . . . . . . . 122
.MEASURE Support for ACPHASENOISE . . . . . . . . . . . . . . . . . . . . . . . 122
Behavioral Noise Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Using Noise Analysis Results as Input Noise Sources . . . . . . . . . . . . . . 124
Syntax Rules for Behavioral Noise Source Files . . . . . . . . . . . . . . . . . . . 126
Noise File Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Power Supply Current and Voltage Noise Sources . . . . . . . . . . . . . . . . . 128
Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

Part III: Periodically-driven AC, Transfer Function, and Noise Analyses

7. Large Signal AC, Transfer Function, and Noise Analyses . . . . . . . . . . . . . 133

Multitone Harmonic Balance AC Analysis (.HBAC) . . . . . . . . . . . . . . . . . . . . . 133
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Tutorial Example - Using Multi-Tone HB and HBAC Analyses for a Mixer 135
Two-tone HB Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
HBAC Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Comparing Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
HBAC Analysis Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
HBAC Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Using the .MEASURE Command with .HBAC . . . . . . . . . . . . . . . . . . . . . 143
Errors and Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Warning Messages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Shooting Newton AC Analysis (.SNAC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

vi
 Contents

SNAC Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
SNAC Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Using the .MEASURE Command with .SNAC . . . . . . . . . . . . . . . . . . . . . 149
Errors and Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Warning Messages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Steady-State Voltage and Current Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
I and V Element Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Phase Noise and Buffer Chains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Multitone Harmonic Balance Noise (.HBNOISE) . . . . . . . . . . . . . . . . . . . . . . . 155
Supported Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
HBNOISE Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
.PRINT and .PROBE Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Measuring HBNOISE Analyses with .MEASURE . . . . . . . . . . . . . . . . . . 160
Optimizing with HBNOISE Measurement. . . . . . . . . . . . . . . . . . . . . . . . . 162
Errors and Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Shooting Newton Noise Analysis (.SNNOISE) . . . . . . . . . . . . . . . . . . . . . . . . 163
Supported Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
SNNOISE Analysis Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Specifying Variant Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
.PRINT and .PROBE Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Measuring SNNOISE Analyses with .MEASURE . . . . . . . . . . . . . . . . . . 169
Multitone Harmonic Balance Transfer Function Analysis (.HBXF). . . . . . . . . . 171
Supported Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
HBXF Test Listing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

vii
Contents

Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

.PRINT and .PROBE Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Using the .MEASURE Command with .HBXF . . . . . . . . . . . . . . . . . . . . . 174
Shooting Newton Transfer Function Analysis (.SNXF). . . . . . . . . . . . . . . . . . . 175
Prerequisites and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
SNXF Test Listing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Using the .MEASURE Command with .SNXF . . . . . . . . . . . . . . . . . . . . . 178
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

8. Large Signal S-parameter Analyses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Frequency Translation S-Parameter (HBLIN) Extraction . . . . . . . . . . . . . . . . . 182
HB Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Port Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Port Element Syntax. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
HBLIN Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
.PRINT and .PROBE Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Large-Signal S-parameter (HBLSP) Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 189
Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
.PRINT and .PROBE Statements. . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Output Data Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
HBLSP Analysis Limitations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

9. Envelope Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

Envelope Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Envelope Analysis Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
Nonautonomous Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

viii
 Contents

Oscillator Analysis Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

Fast Fourier Transform Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Output Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
.PRINT or .PROBE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
.MEASURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Envelope Output Data File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

Part IV: Large Signal Noise Analyses

10. Simulation of Random Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Introduction to Noise Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Characteristics of Random Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Probability Distribution Function versus Power Spectral Density . . . . . . . 206
Multiple Noise Sources in a Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Noise Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Thermal Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Flicker Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
Shot Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Component Noise Models and HSPICE Noise Simulation . . . . . . . . . . . . . . . 217
Element Noise Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
HSPICE Noise Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
RX Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

11. Transient Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

Overview of HSPICE Transient Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . 226
Modeling Frequency-Dependent Noise Sources . . . . . . . . . . . . . . . . . . . . . . 228
Monte Carlo Noise Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Setting up a Monte Carlo Transient Noise Analysis . . . . . . . . . . . . . . . . . . . . . 230
Transient Noise and Monte Carlo Noise Analysis Techniques Available in
HSPICE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Traditional Monte Carlo with Transient Noise Disabled . . . . . . . . . . 230

ix
Contents

Transient Noise Analysis without Monte Carlo (Single Sample). . . . 231

Transient Noise Analysis with Monte Carlo . . . . . . . . . . . . . . . . . . . 231
Both Transient Noise and Traditional Monte Carlo Enabled . . . . . . . 232
Monte Carlo Output Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Monte Carlo Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Controlling Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Setting Noise Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
TIE and Spectrum Jitter Measurements from Transient Noise Results . . . . . . 235
Measure TIE Jitter and Jitter Spectrum Example. . . . . . . . . . . . . . . . . . . 235
Autocorrelation Function Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Phase Noise and Jitter from Transient Noise. . . . . . . . . . . . . . . . . . . . . . . . . . 240
Delay Line Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Using Delay Line Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Phase Detector Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Using the Phase Detector Method . . . . . . . . . . . . . . . . . . . . . . . . . . 242
Phase Noise Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
Jitter Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Power Spectral Density from Transient Noise . . . . . . . . . . . . . . . . . . . . . . . . . 244
Using PSD from Transient Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Stochastic Differential Equation Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
SDE Output Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
SDE Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Error Handling, Error Recovery, Status Reporting . . . . . . . . . . . . . . . . . . . . . . 249
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

12. Periodic Time-Dependent Noise Analysis (.PTDNOISE) . . . . . . . . . . . . . . 251

Introduction to Periodic Time-Dependent Noise Analysis . . . . . . . . . . . . . . . . 251
PTDNOISE Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
PTDNOISE Output Syntax and File Format. . . . . . . . . . . . . . . . . . . . . . . . . . . 254
Output File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
.MEASURE Syntax and File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Errors and Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Troubleshooting .PTDNOISE Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

x
 Contents

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

Part V: Behavioral Modeling Applications

13. Behavioral Modeling of Nonlinear Circuits . . . . . . . . . . . . . . . . . . . . . . . . . 263

Behavioral Design Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Using Behavioral Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Controlled Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
Voltage and Current Controlled Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Modeling with Digital Behavioral Components. . . . . . . . . . . . . . . . . . . . . . . . . 267
Behavioral AND and NAND Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Behavioral D-Latch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Behavioral Double-Edge Triggered Flip-Flop . . . . . . . . . . . . . . . . . . . . . . 270
Calibrating Digital Behavioral Components . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
Building Behavioral Lookup Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
Subcircuit Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Behavioral N-Channel MOSFET . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
Creating a Behavioral Inverter Lookup Table . . . . . . . . . . . . . . . . . . 274
Optimizing Behavioral CMOS Inverters . . . . . . . . . . . . . . . . . . . . . . . . . . 276
Optimizing Behavioral Ring Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . 278
Analog Behavioral Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Behavioral Differentiator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Behavioral Amplitude Modulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Behavioral Data Sampler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Op-Amps, Comparators, and Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
741 Op-Amp from Controlled Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
Inverting Comparator with Hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Voltage-Controlled Oscillator (VCO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
LC Oscillator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
Phase-Locked Loops (PLL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Phase Detector, with Multi-Input NAND Gates. . . . . . . . . . . . . . . . . . . . . 293
Phase Locked Loop Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298

xi
Contents

14. Laplace Transform Modeling of Filters and Networks . . . . . . . . . . . . . . . . 299

Transient Modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Using G- and E-elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Laplace Transform Function Call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
G- and E-element Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Laplace Band-Reject Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Laplace Low-Pass Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Circular Convolution Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
30-Degree Phase Shift Circuit File . . . . . . . . . . . . . . . . . . . . . . . . . . 307
Laplace and Pole-Zero Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Laplace Transform (LAPLACE) Function . . . . . . . . . . . . . . . . . . . . . . . . . 309
General Form of the Transfer Function. . . . . . . . . . . . . . . . . . . . . . . 309
Finding the Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
Determining the Laplace Coefficients . . . . . . . . . . . . . . . . . . . . . . . 313
Laplace Transform POLE (Pole/Zero) Function . . . . . . . . . . . . . . . . . . . . 317
POLE Function Call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
General Form of the Transfer Function. . . . . . . . . . . . . . . . . . . . . . . 317
Reduced Form of the Transfer Function . . . . . . . . . . . . . . . . . . . . . . 318
RC Line Modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
AWE Transfer Function Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Y-parameter Line Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
Comparison of Circuit and Pole/Zero Models . . . . . . . . . . . . . . . . . . . . . 329
Simulation Time Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Modeling Switched Capacitor Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Switched Capacitor Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Switched Capacitor Filter Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
Input File for Switched Capacitor Filter . . . . . . . . . . . . . . . . . . . . . . . . . . 334
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

xii
 About This Guide

This user guide focuses on advanced analog analysis and modeling and
concentrates on advanced analog circuit characterization, noise modeling and
analysis, and behavioral analysis including Verilog-A.

Related Products and Trademarks

This manual refers to the following products:
Synopsys Custom WaveView™
Synopsys SolvNet® support site

Topics in This Document

This manual contains the chapters described in the following table. For
descriptions of the other manuals in the HSPICE documentation set, see The
HSPICE Documentation Set.

Chapter Description

Chapter 1, HSPICE Advanced Describes how to use HSPICE advanced analog analyses,
Analog Features take advantage of its specialized features, and use the
Custom WaveView tool; introduces the HSPICE solutions for
noise analysis.

Chapter 2, Harmonic Balance Describes how to use harmonic balance analysis for
Analysis frequency-driven and steady-state analysis.

Chapter 3, Shooting Newton Describes the HSPICE advanced analog steady-state time
Analysis domain analysis based on a Shooting-Newton algorithm.

Chapter 4, Oscillator Analysis Describes how to use the HSPICE advanced analog
functions to perform oscillator and phase noise analysis on
oscillator circuits.

HSPICE® User Guide: Advanced Analog Simulation and Analysis xiii

N-2017.12
About This Guide
Topics in This Document

Chapter Description

Chapter 5, Phase Noise Analysis Describes phase noise analysis.

Chapter 6, Behavioral PLL Noise Describes behavioral PLL noise analysis.

Analysis (.ACPHASENOISE)

Chapter 7, Large Signal AC, Describes how to use both harmonic balance-based and
Transfer Function, and Noise Shooting Newton-based AC analysis and transfer function
Analyses analyses, as well as nonlinear, steady-state noise analysis.

Chapter 8, Large Signal S- Describes how to do frequency translation and large-signal

parameter Analyses S-parameter extraction, as well as noise parameter
calculation.

Chapter 9, Envelope Analysis Describes how to use envelope simulation.

Chapter 10, Simulation of Describes the characteristics of random signals, types of

Random Noise noise, component noise models, and noise simulation in
HSPICE.

Chapter 11, Transient Noise Describes the HSPICE solutions to perform transient noise
Analysis analysis and compute noise statistics and their variation over
time for circuits driven with non-periodic waveforms.

Chapter 12, Periodic Time- Describes periodic time-dependent noise analysis.

Dependent Noise Analysis
(.PTDNOISE)

Chapter 13, Behavioral Modeling Describes how to create and use behavioral models.
of Nonlinear Circuits

Chapter 14, Laplace Transform Describes modeling filters and networks, including Laplace
Modeling of Filters and Networks transforms.

xiv HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 About This Guide
The HSPICE Documentation Set

The HSPICE Documentation Set

This manual is a part of the HSPICE documentation set, which includes the
following manuals:

Manual Description

HSPICE® Quick Start Guide Provides information to quickly get you started with using the
HSPICE tool for simulation and analysis of your circuit designs.
See also Topics in This Document in the HSPICE Quick Start
Guide.

HSPICE® User Guide: Basic Describes how to use HSPICE to simulate and analyze your
Simulation and Analysis circuit designs, and includes simulation applications. This is the
main HSPICE user guide. See also Topics in This Document in
the HSPICE User Guide: Basic Simulation and Analysis.

HSPICE® User Guide: Describes the syntax for the basic elements of a circuit netlist in
Elements HSPICE, descriptions of each of the element keywords, and
examples of common usage for each element. See also Topics
in This Document in the HSPICE User Guide: Elements.

HSPICE® User Guide: Signal Describes how to use the HSPICE simulator to achieve and
Integrity Modeling and maintain signal integrity in your chip design. See also Topics in
Analysis This Document in the HSPICE User Guide: Signal Integrity
Modeling and Analysis.

HSPICE® User Guide: This guide. Describes how to use the special set of analysis and
Advanced Analog Simulation design capabilities added to HSPICE to support RF and high-
and Analysis speed circuit design. See also Topics in This Document.

HSPICE® Reference Manual: Describes the individual HSPICE commands you can use to
Commands and Control simulate and analyze your circuit designs. See also Topics in
Options This Document in the HSPICE Reference Manual: Commands
and Control Options.

HSPICE® Reference Manual: Describes standard models you can use when simulating your
Device Models circuit designs in HSPICE, including passive devices, diodes,
JFET and MESFET devices, and BJT devices. See also Topics
in This Document in the HSPICE Reference Manual: Device
Models.

HSPICE® User Guide: Advanced Analog Simulation and Analysis xv

N-2017.12
About This Guide
Conventions

Manual Description

HSPICE® Reference Manual: Describes the available MOSFET models you can use when
MOSFET Models simulating your circuit designs in HSPICE. See also Topics in
This Document in the HSPICE Reference Manual: MOSFET
Models.

HSPICE® Integration to Describes use of the HSPICE simulator integration to the

Cadence® Virtuoso® Analog Cadence tool. See also Topics in This Document in the HSPICE
Design Environment User Integration to Cadence Virtuoso Analog Design Environment.
Guide

Conventions
This manual follows these typographical conventions in Synopsys HSPICE®
documentation.

Convention Description

Courier Indicates command syntax.

Italic Indicates a user-defined value, such as object_name.

Bold Indicates user input—text you type verbatim—in syntax and examples.
For a graphical user interface, Bold indicates a GUI element such as a
button, menu, field, or other control.

[ ] Denotes optional parameters, such as:

write_file [-f filename]

( ) When shown, the parentheses ( ) are part of the syntax. For example:
+ LISTFREQ=(1k 100k 10meg)

... Indicates that parameters can be repeated as many times as necessary:

pin1 pin2 ... pinN

| Indicates a choice among alternatives, such as

low | medium | high

+ Indicates a continuation of a command line.

/ Indicates levels of directory structure.

xvi HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 About This Guide
Customer Support

Convention Description

Edit > Copy Indicates a path to a menu command, such as opening the Edit menu and
choosing Copy.

Control-c Indicates a keyboard combination, such as holding down the Control key
and pressing c.

Customer Support
Customer support is available through the Synopsys SolvNet customer support
website and by contacting the Synopsys support center.

Accessing SolvNet
The SolvNet support site includes an electronic knowledge base of technical
articles and answers to frequently asked questions about Synopsys tools. The
site also gives you access to a wide range of Synopsys online services, which
include downloading software, viewing documentation, and entering a call to
the Support Center.
To access the SolvNet site:
1. Go to the web page at https://solvnet.synopsys.com.
2. If prompted, enter your user name and password. (If you do not have a
Synopsys user name and password, follow the instructions to register.)
If you need help using the site, click Help on the menu bar.

Contacting Synopsys Support

If you have problems, questions, or suggestions, you can contact Synopsys
support in the following ways:
■
Go to the Synopsys Global Support site on synopsys.com. There you can
find e-mail addresses and telephone numbers for Synopsys support centers
throughout the world.
■
Go to either the Synopsys SolvNet site or the Synopsys Global Support site
and open a case online (Synopsys user name and password required).

HSPICE® User Guide: Advanced Analog Simulation and Analysis xvii

N-2017.12
About This Guide
Customer Support

xviii HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 1
1 HSPICE Advanced Analog Features

Describes how to use HSPICE advanced analog analyses, take advantage of

its specialized features, and use the Custom WaveView tool; introduces the
HSPICE solutions for noise analysis.

HSPICE advanced analog analyses is a set of analysis and design capabilities

that support the design of advanced analog and high-speed circuits. HSPICE
advanced analog analyses includes several modeling, simulation, and
measurement additions that augment the ultimate-accuracy analog circuit
simulation capabilities.
HSPICE advanced analog analyses accepts a netlist file from the standard
input and delivers the ASCII text simulation results to HTML or to the standard
output. Standard error output reports the error and warning messages.
These following sections cover these topics:
■
Introduction to HSPICE Advanced Analog Analyses
■ Use of Example Syntax

Introduction to HSPICE Advanced Analog Analyses

The following sections discuss the advanced analog features:
■ HSPICE Advanced Analog Analyses
■
HSPICE Advanced Analog Features
■
Recommended Analysis Method for Typical RF Design Blocks
■ HSPICE Advanced Analog Output Files
■
Using Custom WaveView

HSPICE® User Guide: Advanced Analog Simulation and Analysis 1

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

■
Creating a Configuration File
■
Using Wildcards
■
Limiting Output Data Size
■
Generating Measurement Output Files
■
Advanced Analog Demonstration Input Files

HSPICE Advanced Analog Analyses

HSPICE advanced analog analyses provide several analyses that support the
simulation and analysis of radio-frequency integrated circuits (RFICs). The
advanced analog analyses include:
■
Harmonic Balance Analysis
■
Shooting Newton Analysis and Shooting Newton with Fourier Transform
(.SNFT)
■
Oscillator Analysis (.HBOSC and .SNOSC)
■
Shooting Newton Oscillator Analysis (.SNOSC)
■
Large Signal AC, Transfer Function, and Noise Analyses and Multitone
Harmonic Balance AC Analysis (.HBAC)
■
Shooting Newton AC Analysis (.SNAC)
■
Large Signal AC, Transfer Function, and Noise Analyses
■
Shooting Newton Noise Analysis (.SNNOISE)
■ Multitone Harmonic Balance Transfer Function Analysis (.HBXF)
■
Shooting Newton Transfer Function Analysis (.SNXF)
■
Frequency Translation S-Parameter (HBLIN) Extraction
■
Envelope Analysis
■
Transient Noise Analysis

Important: You can enable the analytical derivative computation of

expression-based element evaluations. You can enable this
in all advanced analog analyses, for extensive accuracy, by
using the .OPTION EQN_ANALYTICAL_DERIV described
in the HSPICE Reference Manual: Commands and Control
Options.

2 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

HSPICE Advanced Analog Features

This section briefly introduces the features of both the simulation engine and
the waveform display tool.
■
Steady-state frequency-domain analyses for linear and nonlinear circuits.
■
The.HBLSP command invokes periodically-driven nonlinear circuit
analyses for power-dependent S parameters.
■
Harmonic Balance (.HB) analysis by using Direct and Krylov solvers.
The.HB command invokes the single- and multi-tone Harmonic Balance
algorithm for periodic steady-state analysis.
■
TRANFORHB element parameter to recognize V/I sources that include SIN
and PULSE transient descriptions as well as PWL and VMRF sources.
■
Harmonic balance-based periodic AC analysis. The .HBAC command
invokes periodic AC analysis for analyzing small-signal perturbations on
circuits that operate in a large-signal periodic steady-state.
■
Harmonic Balance-based Periodic Noise analysis (.HBNOISE) for noise
analysis of periodically modulated circuits; includes stationary,
cyclo-stationary, and frequency-dependent noise effects.
■
Autonomous Harmonic Balance analysis. The.HBOSC command invokes
the multi-tone, oscillator-capable Harmonic Balance algorithm for periodic
steady-state analysis.
■
Perturbation analysis for Oscillator Phase Noise. The .HBAC command
invokes phase periodic AC noise for oscillators circuits that operate in a
large-signal steady-state.
■
Oscillator phase noise analysis, including both a nonlinear perturbation
method and a PAC method; includes stationary, cyclo-stationary, frequency-
dependent, and correlated noise effects.
■
Frequency translation S-parameter and noise figure extraction with
the .HBLIN command.
■
Envelope analysis. The.ENV command invokes standard envelope
simulation. The .ENVOSC command invokes envelope startup simulation.
The.ENVFFT command invokes envelope Fast Fourier Transform
simulation.
■
.OPTION HBTRANINIT, HBTRANPTS, and HBTRANSTEP, to use transient
analysis to initialize all state variables for HB analysis.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 3

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

■
Calculation of the transfer function from an arbitrary source and harmonic in
the circuit to a designated output with the .HBXF command.
■ .OPTION SIM_ACCURACY provides simplified accuracy control for all
simulations while .OPTION SIM_ORDER and SIM_TRAP improve transient
analysis simulation controls.
■
DSPF Flow for fast analysis by using parasitic data from layout.
■
Shooting Newton steady-state time domain analysis; the Shooting Newton
algorithm provides functionality to support the following commands: .SN,
.SNAC .SNFT, .SNNOISE, .SNOSC, and .SNXF.
■
Periodic Time-Dependent Noise Analysis (.PTDNOISE) calculates the
noise spectrum and the total noise at a point in time. This analysis
determines jitter in a digital threshold circuit from the total noise and the
digital signal slew rate.
■
HSPICE advanced analog analyses supports the ISUB syntax with the
exception of wildcard support with the “?” sign. For example, HSPICE
advanced analog analyses does not support isub(x1.a?).
■
HSPICE advanced analog analyses supports case sensitivity.
HSPICE advanced analog analyses also includes the following measurement
capabilities:
■
1dB compression point.
■
Intercept points (for example, IP2, IP3).
■
Mixer conversion gain and noise figure.
■
VCO output spectrum.
■
Oscillator phase noise.
■
Options simplify specifying levels of accuracy. As a result, HSPICE provides
effective simulation solutions for advanced analog, high-speed, and PCB
signal integrity circuit challenges.
■
Verilog-A is supported for all advanced analog analyses.
Standard restrictions for Verilog-A in periodic steady-state analysis are the
same as other advanced analog simulators that use Verilog-A. For example:
• Verilog-A modules that are time-dependent are illegal for HB or SN
unless the time dependence is periodic with a period that matches the
HB or SN setup.

4 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

• Verilog-A modules with “internal states” may not work correctly in HB or

SN because the engine cannot track the internal state. So, HB or SN
may accept convergence to a periodic steady-state even though the
internal state may not be in periodic steady state.
• Some event-driven constructs in Verilog-A may not be compatible with
HB.
■
For netlist input guidelines, see Input Netlist and Data Entry in the HSPICE
User Guide: Basic Simulation and Analysis.
■
For information on the use of parameters and functions, see Parameters
and Functions in the HSPICE User Guide: Basic Simulation and Analysis.
■
For information on the use of Monte Carlo sweeps, see Variation Analysis in
the HSPICE User Guide: Basic Simulation and Analysis.

Recommended Analysis Method for Typical RF Design

Blocks
Design SN Analysis HB Analysis

Oscillators Yes Yes

Crystal oscillators - Yes

Frequency dividers Yes -

Switch cap circuits Yes -

Mixers - Yes

Charge pumps Yes -

LNA Yes Yes

As a general rule of thumb, consider using:

■ Shooting Newton analysis for highly nonlinear circuits or circuits having
waveforms with sharp transitions.
■
Harmonic Balance analysis for the other types of circuits.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 5

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

HSPICE Advanced Analog Output Files

The following table shows the output file extensions that HSPICE advanced
analog analyses produce. The base file name of each output file is the same as
the base name of the input netlist file. The # at the end of each file extension
represents the .ALTER run from which the file came.
In general, text output from .PRINT commands is for users, while binary output
from .PROBE or .OPTION POST is the input for the Custom WaveView tool.

Command Description Text Output Output for Custom

WaveView

.ENV Envelope analysis .printev# .ev#

.ENVFFT Envelope FFT (none) .fe#

.ENVOSC Oscillator Startup .printev# .ev#

.HB Harmonic Balance .printhb# .hb#

.HBAC Harmonic Balance AC .printhb# .hb#

.HBLIN Harmonic Balance .PRINT output: .printhl# .PROBE output: .hl#

Linear Analysis S-param output: .SnP S-param output: .SnP

.HBLSP .HBLSP large-signal .PRINT output: .printls# .PROBE output: .ls#

S-param output: .p2d S-param output: .p2d

.HBLSP small-signal .PRINT output: .printss# .PROBE output: .ss#

S/noise output: .S2P# S/noise output: .S2P#

.HBNOISE HBAC noise .printsnpn# .pn#

.HBOSC Harmonic Balance OSC .printhb# .hb#

.HBXF Transfer Functions .printxf# .xf#

.PHASENOISE Phase Noise (SNOSC) .printsnpn# .pn#

Phase Noise (HBOSC) .printpn# .pn#

Jitter .printjt# .jt#

.PTDNOISE Periodic Time .printptn# .ptn#

Dependent Noise (HB)

Periodic Time .printsnptn# .snptn#

Dependent Noise (SN)

6 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

Command Description Text Output Output for Custom

WaveView

.SN Shooting Newton .printsn# .sn#

Analysis

.SNAC Shooting Newton AC .printsnac# .snac#

.SNNOISE Shooting Newton Noise .printsnpn# .snpn#

.SNOSC Shooting Newton OSC .printsn# .sn#

.SNXF Shooting Newton .printsnxf# .snxf#

Transfer Function

.TRANNOISE Transient Noise .printtrpn# .trpn#

Jitter .printtrjt# .trjt#

Autocorrelation Function none .trzn#

Using Custom WaveView

Synopsys Custom WaveView supports viewing and processing of HSPICE
output files. This section presents a basic overview of how to use the tool.
1. To start the Custom WaveView tool, type wv on the UNIX/Linux command
line.
2. Choose File > Import Waveform File (or press CTRL+O) to open the Open
Waveform Files dialog box.
3. Use the File Filters to limit the file names to Waveform Files. The preceding
table lists the HSPICE advanced analog file types. When you open a file, its
contents appear in the file browser. The file browser lists all open plot files.
Click on the '+' near a waveform file name, to display the hierarchy of the
waveform. Click on the top level or any hierarchy level to display the contents
of the waveform file to appear in the signal browser. To plot one of the
signals listed in the waveform, you can either double-click the signal label or
select, drag, and drop the signal label to WaveView.
4. To create a panel, use the Panel > New menu and select the panel type:
X-Y, Smith Chart, or Polar Plot. You can also use the panel icon in the tool
bar to create new panels.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 7

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

5. To create a new chart, use the File > New menu. Select either XY Graph,
Smith Chart, or Polar Chart. You can also use the first three icons in the
tool bar to create new chart windows.
6. Use the Custom WaveView tool bar to change how signals look, to delete
signals, and to group or ungroup signals.
7. Right-click on a frequency-domain signal name and use the To Time
Domain command to convert the histogram waveform (for example, from
a .hb0 file) to a time domain waveform.
8. Configure the axis scale and grid by right-clicking a horizontal or vertical axis
and selecting the desired scale or grid from the context-sensitive menu.
9. Zoom in and out, using the zoom icons on the tool bar, or use the mouse
cursor to select an area directly on the waveform.
10. You can use dynamic meters to see the precise value of the signal at
different points. From the menu, select Tools > Dynamic Meter or use the
Dynamic Meter icon in the tool bar. Select and configure the desired
dynamic meter. You can then move the meter to the desired location on the
selected signal.
11. To use the measurement tools, choose Tools > Measurement. You can use
the following advanced analog options available under the All tab of the
Measurement Tool window:
• 1db compression point (P1dB).
• 2nd order intercept point (IP2).
• 3rd order intercept point and spurious free dynamic range (IP3/SFDR).

Creating a Configuration File

You can create a configuration file, called .hspicerf, to customize your HSPICE
advanced analog simulation. HSPICE first searches for .hspicerf in your current
working directory and then in your home directory as defined by $HOME.

8 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

Following are the configuration options available in HSPICE:

Keyword Description Example

flush_waveform Flushes a waveform. If you do not specify a percentage, flush_waveform

then the default value is 20%. percent%

ground_floating_node Uses .IC statements to set floating nodes in a circuit to ground_floating_

ground. You can select three options for grounding floating node 1
nodes:
■
If set to 1, grounds only floating nodes (gates, bulk,
control nodes, and non-rail bulk) that the .IC set
includes.
■
If set to 2, adds unconnected terminals to this set.
■ If set to 3, uses .IC statements to ground all floating
nodes, including dangling terminals.

hier_delimiter Changes the delimiter for subcircuit hierarchies from “.” to hier_delimiter /
the specified symbol.

html Stores all HSPICE output in HTML format. htmlhspicerf test

This example creates a file
named test.html in the
current directory.

integer_node Removes leading zeros from node names. For example, integer_node
HSPICE considers 0002 and 2 to be the same node.

Without this keyword, 0002 and 2 are two separate nodes.

max_waveform_size Automatically limits the waveform file size. max_waveform_

■ size 2000000000
If the number is less than 5000, HSPICE resets it to 2G.
■ If you do not set the number, HSPICE uses the default,
2G.
■
If you do not set the line, the file size has no limit.

negative_td Allows negative time delay input in pwl (piecewise linear If you do not set
with repeat), pl (piecewise linear), exp (exponential, rising negative_td, the
time delay only), sin (damped sinusoidal), pulse negative time delay defaults
(trapezoidal pulse), and am (amplitude modulation) to zero.
formats.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 9

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

Keyword Description Example

port_element_ Allows the alternate Port element definition. A Port element port_element_
voltage_ matchload consists of a voltage source in series with a resistor. voltage_
For the explanation that follows, let the user-specified DC, matchload
AC, or transient value of the Port element be V, and let the
voltage across the overall port element be Vp.
By default, HSPICE advanced analog analyses set the
internal voltage source value to V. The value of Vp is lower
than V, depending on the internal impedance and the
network's input impedance.
With the alternate definition, the internal voltage source
value is adjusts to 2*V, so that Vp=V when the Port
element's impedance matches the network input
impedance. The actual value of Vp still depends on the port
and network impedances.

rcxt_divider Defines the hierarchy delimiter in the active nodes file in rcxt_divider /
RCXT format.

skip_nrd_nrs Directs HSPICE advanced analog analyses to consider skip_nrd_nrs

transistors with matching geometries (except for NRD and
NRS) as identical for pre-characterization purposes.

unit_atto Activates detection of the “atto.” unit. Otherwise, HSPICE unit_atto

advanced analog analyses assumes that “a” represents
“amperes.”

v_supply Changes the default voltage supply range for v_supply 3

characterization.

wildcard_left_range Begins range expression. wildcard_left_

range [

wildcard_match_all Matches any group of characters. wildcard_match_

all *

wildcard_match_one Matches any single character. wildcard_match_

one ?

wildcard_right_range Ends range expression. wildcard_right_

range ]

Note: For more information about wildcards, see Using Wildcards.

10 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

This section covers the following topics:

■
Inserting Comments in a .hspicerf File

Inserting Comments in a .hspicerf File

To insert comments into your .hspicerf file, include a number sign character
(#) as the first character in a line. For example, this configuration file shows how
to use comments in a .hspicerf file:
# sample configuration file
# the next line of code changes the delimiter
# for subcircuit hierarchies from "," to "^"
hier_delimiter ^
# the next line of code matches any groups of "*" characters
wildcard_match_all *
# the next line of code matches one "?" character
wildcard_match_one ?
# the next line of code begins the range expression with
# the "[" character
wildcard_left_range [
# the next line of code ends the range expression with
# the "]" character
wildcard_right_range ]

Using Wildcards
You can use wildcards to match node names. HSPICE advanced analog uses
wildcards somewhat differently than standard HSPICE.
Before you use wildcards, you must define the wildcard configuration in
a .hspicerf file. For example, you can define the following wildcards in
the .hspicerf file:
file .hspicerf
wildcard_match_one ?
wildcard_match_all *
wildcard_left_range [
wildcard_right_range ]

The .PRINT, .PROBE, .LPRINT, and .CHECK statements support wildcards in

HSPICE.
For more information about using wildcards in an HSPICE configuration file,
see Using Wildcards in PRINT and PROBE Statements in the HSPICE User
Guide: Basic Simulation and Analysis.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 11

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

Limiting Output Data Size

For multi-million transistor simulations, an unrestricted waveform file can grow
several gigabytes in size. The file becomes unreadable in some waveform
viewers, and requires excessive space on the hard drive.
This section describes options that limit the number of nodes output to the
waveform file, to reduce the file size. HSPICE supports the following options to
control the output:

Control Option Description

.OPTION SIM_POSTTOP Use this option to limit the data written to your waveform file to data from
only the top n level nodes. This option outputs instances up to n levels
deep.
Note: To enable the waveform display interface, you also need the
POST option.

.OPTION SIM_POSTSKIP Use this option to have the SIM_POSTTOP option skip any instances
and their children that the subckt_definition defines.

.OPTION SIM_POSTAT Use this option to limit the waveform output to only the nodes in the
specified subcircuit instance.
This option can operate in conjunction with the SIM_POSTTOP option
and when present, has precedence over the SIM_POSTSKIP option.

.OPTION SIM_POSTDOWN Use this option to include an instance and all children of that instance
in the output.
This option can operate in conjunction with the SIM_POSTTOP option
and when present, has precedence over the SIM_POSTSKIP option.

.OPTION SIM_POSTSCOPE Use this option to specify the signal types to probe from within a scope.

Generating Measurement Output Files

You can make measurements with the .MEASURE statement when using
HSPICE advanced analog analyses.
The results of the .MEASURE statements appear in a file with one of the
following filename extensions:
■
.mb# for measurements in HB analysis
■ .mp# for measurements in HBNOISE and SNNOISE analysis

12 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 1: HSPICE Advanced Analog Features
Introduction to HSPICE Advanced Analog Analyses

■
.mpn# for measurements from PHASENOISE analysis when using HB to
obtain the steady state solution
■ .msnpn# for measurements from PHASENOISE analysis when using SN to
obtain the steady state solution
■
.msnptn# for measurements in PTDNOISE analysis
For more information about .MEASURE statements, see HSPICE Netlist
Commands in the HSPICE Reference Manual: Commands and Control
Options.

Advanced Analog Demonstration Input Files

The following is a listing of shipped demonstration files for illustrating the
HSPICE advanced analog analyses functionality. All of these example files are
available at:
$installdir/demo/hspice/rf_examples

File Name Description

acpr.sp Envelope simulation example

bjt.inc Transistor model library used by osc.sp

clockbuf4.sp Transient noise analysis example

clockbuf4sn.sp Shooting Newton-based analysis of a clock buffer circuit example

cmos49_model.inc Transistor model library used by example circuits

cmos90nmWflicker.lib Transistor model library used by phasefreqdet.sp

gpsvco.sp Oscillator and Phase Noise analysis example

gsmlna.sp LNA Linear analysis example

gsmlnaIP3_A.sp 3rd order intercept point example

mix_hb.sp Mixer HB analysis example

mix_hbac.sp MIxer HBAC analysis example

mix_snac.sp Mixer Shooting Newton AC example

mix_tran.sp Mixer transient analysis example

HSPICE® User Guide: Advanced Analog Simulation and Analysis 13

N-2017.12
Chapter 1: HSPICE Advanced Analog Features
Use of Example Syntax

File Name Description

osc.sp Oscillator tuning curve and phase noise analysis example

pa.sp Power amplifier HB analysis example

pfdcpGain.sp Shooting Newton analysis example

phasefreqdet.sp Shooting Newton and noise analysis example

rcInvPTDnoise.sp Periodic Time-Dependent Noise analysis (.PTDNOISE) example

ringoscSN.sp Shooting Newton and Phase Noise analysis example

tsmc018.m Transistor model library used by ringoscSN.sp

Use of Example Syntax

To copy and paste proven syntax, use the demonstration files shipped with your
installation of HSPICE (see Listing of Demonstration Input Files in HSPICE
User Guide: Basic Simulation and Analysis). Attempting to copy and paste from
the book or help documentation may present unexpected results, as text used
in formatting may include hidden characters and white space for visual clarity.

14 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Part 1: Periodic Steady-State Analysis

Part 1 presents the following chapters /topics:

■
Chapter 2, Harmonic Balance Analysis
■ Chapter 3, Shooting Newton Analysis

HSPICE® User Guide: Advanced Analog Simulation and Analysis 15

N-2017.12
16 HSPICE® User Guide: Advanced Analog Simulation and Analysis
N-2017.12
 2
Harmonic Balance Analysis
2

Describes how to use harmonic balance analysis for frequency-driven and

steady-state analysis.

This chapter provides an introduction to the Harmonic Balance (.HB) analysis

command. You can use steady-state analysis on a circuit if it contains only DC
and periodic sources. These analyses assume that all “start-up” transients
have completely died out with only the steady-state response remaining. These
analyses treat sources that are not periodic or DC as zero-valued.
The following sections discuss these topics:
■
Tutorial Examples Harmonic Balance Analysis
■
Harmonic Balance Analysis
■ Troubleshooting Convergence Issues with Harmonic Balance Analysis
■
Phase Differences Between HB and SIN Sources
■
Optimization
■ Optimizing with HB Measurements
■
References

Tutorial Examples Harmonic Balance Analysis

The following sections present these tutorial examples:
■
Example 1: Using HB Analysis for a Power Amplifier
■
Example 2: Using HB Analysis for a Low Noise Amplifier

HSPICE® User Guide: Advanced Analog Simulation and Analysis 17

N-2017.12
Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Example 1: Using HB Analysis for a Power Amplifier

The .HB command computes periodic steady-state solutions of circuits. This
analysis uses the Harmonic Balance (HB) technique for computing such
solutions in the frequency domain. The circuit can be driven by a voltage,
power, or current source, or it may be an autonomous oscillator. The HB
algorithm represents the circuit’s voltage and current waveforms as a Fourier
series; that is, a series of sinusoidal waveforms.
To set up a periodic steady-state analysis, the HSPICE input netlist must
contain:
■
An .HB command to activate the analysis. The .HB command specifies the
base frequency (or frequencies, also called tones) for the analysis, and the
number of harmonics to use for each tone. The .HB command can specify
base tones so that the circuit solution is represented as a multi-dimensional
Fourier series. The number of terms in the series are determined by the
number of harmonics; more harmonics result in higher accuracy, but also
longer simulation times and higher memory usage.
■
One or more signal sources for driving the circuit in HB analysis, if the circuit
is driven. In the case of autonomous oscillator analysis, no signal source is
required. Signal sources are specified using the HB keyword in the voltage
or current source syntax. Power sources are specified by setting the power
switch on the voltage/current sources to 1; in this case, the source value is
treated as a power value in Watts, instead of a voltage or current.
Optionally, the netlist can also contain a set of control options for optimizing the
HB analysis performance.
The following example shows how to set up an HB analysis on an NMOS Class
C Power Amplifier. The example compares transient analysis results to the HB
results.

18 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Figure 1 Power Amplifier

The following netlist performs both a transient and an HB analysis of the

amplifier driven by a sinusoidal input waveform. The accurate option is set to
ensure sufficient number of time points for comparison with HB. This example
is included with the HSPICE advanced analog analyses distribution as pa.sp
and is available in directory $installdir/demo/hspice/rf_examples/.
.options POST accurate
.param f0=950e6 PI=3.1415926 Ld=2e-9 Rload=5 Vin=3.0
.param Lin=0.1n Vdd=2 Cd='1.0/(4*PI*PI*f0*f0*Ld)'
M1 drain gt 0 0 CMOSN L=0.35u W=50u AS=100p AD=100p
PS=104u PD=104u M=80
Ls in gt Lin $ gate tuning
Ld drain vdd Ld $ drain tuning
Cd drain 0 Cd
Cb drain out INFINITY $ DC block
Rload out 0 Rload
Vdd vdd 0 DC Vdd
Vrf1 in 0 DC 'Vin/2.0'
+ SIN ('Vin/2' 'Vin/2' 'f0' 0 0 90)
+ HB 'Vin/2' 0.0 1 1
.hb tones=f0 nharms=10
.tran 10p 10n
.probe hb p(Rload)
.probe tran p(Rload)
.include cmos49_model.inc

HSPICE® User Guide: Advanced Analog Simulation and Analysis 19

N-2017.12
Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

.end

An HB analysis uses the following:

■
An .HB command:
.hb tones=f0 nharms=10

For a single-tone analysis with base frequency 950 MHz and 10 harmonics.
■
The HB source in Vrf1:
HB ‘Vin/2’ 0.0 1 1.

This creates a sinusoidal waveform matching the transient analysis one.

The amplitude is Vin/2=1.5 V, and it applies to the first harmonic of the first
tone, 950 MHz.
■
A .PROBE command for plotting the output power:
.probe hb p(Rload)

To run this netlist, type the following command:

hspice pa.sp

This produces two output files named pa.tr0 and pa.hb0, containing the
transient and HB output, respectively. To view and compare the output:
1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the pa.tr0 and pa.hb0 files
from the Open: Waveform Files dialog box.
3. Select the v(out) signal from pa.hb0 in the signal browser. Double-click on
the signal name, or drag and drop the signal in the waveform. The histogram
shows lines at 950MHz, and multiples thereof, up to 9.5GHz.

20 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

4. In the waveform, right click in the name area of the panel containing the
signal v(out), left-click on the waveform label for v(out) from the pa.hb0 file.
From the Panel menu, choose Signal 'v(out)' > To Time-Domain.
5. In the Convert to Time Domain window, change the X-End (sec) value to
10n.
6. Click OK to accept the settings.
7. The new waveform shows a new time domain waveform named IFT.0|v(out).

HSPICE® User Guide: Advanced Analog Simulation and Analysis 21

N-2017.12
Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

8. Select the v(out) signal from pa.tr0 in the signal browser. Drag and drop
the signal in the waveform containing IFT.0|v(out). This should overlay the
IFT.0|v(out) and v(out) signals on the same waveform. Zoom into the
transitions to see the slight differences between the waveforms.

Example 2: Using HB Analysis for a Low Noise Amplifier

This example performs a simulation on a Low Noise Amplifier circuit using two
closely-spaced steady-state tones to study the compression and the third-order
distortion properties of the amplifier. The example file gsmlnaIP3_A.sp is
located at: $installdir/demo/hspice/rf_examples/.

See Figure 2 for the schematic view.

22 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Figure 2 Schematic showing instantiation for Low Noise Amplifier

HSPICE® User Guide: Advanced Analog Simulation and Analysis 23

N-2017.12
Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Example 1 Low Noise Amplifier Netlist: Part 1

** NMOS 0.25um Cascode LNA for GSM applications
** setup for s-parameter and noise parameter measurements
.temp 27
.options post
.param Vdd=2.3
**
** Cascode LNA tuned for operation near 1 GHz
**
M1 _n4 _n3 _n5 _n5 CMOSN l=0.25u w=7.5u
+ as=15p ad=15p ps=19u pd=19u m=80
M2 _n6 _n1 _n4 _n4 CMOSN l=0.25u w=7.5u
+ as=15p ad=15p ps=19u pd=19u m=80
M3 rfo _n6 gnd gnd CMOSN l=0.25u w=7.5u
+ as=15p ad=15p ps=19u pd=19u m=40
r1 _vdd _n6 400
l1 _n5 gnd l=0.9nH
l2 rfin _n3 l=13nH
lchk rfin rfinb INFINITY
cblk rfin rfind INFINITY
vvb _n1 gnd dc=1.19 $ bias for common base device
vinb rfinb gnd dc=0.595
vvdd _vdd gnd dc=Vdd
rfb rfo _n6 120 $ feedback
**
**
** Two-tone input source (DC blocked at this time)
**
Vin rfind gnd dc=0 power=1 z0=50 $ 50 Ohm src
+ HB Pin 0 1 1 $ tone 1
+ HB Pin 0 1 2 $ tone 2
Rload rfo _vdd R=255
**
** HB test bench to measure IP3 and IP2
.HB tones=900MEG, 910MEG nharms=11 11 intmodmax=7
+ SWEEP Pin dec 10 1e-8 1e-3
.print HB P(Rload) P(Rload)[1,0] P(Rload) [2,0] P(Rload)[2,-1]
.probe HB P(Rload) P(Rload)[1,0] P(Rload) [2,0] P(Rload)[2,-1]
**
**

24 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Example 2 Low Noise Amplifier Netlist: Part 2

*
.meas hb fund1_mag max v(rfo) [1,0]
.meas hb fund2_mag max v(rfo) [0,1]
.meas hb harm1_mag max v(rfo) [2,-1]
.meas hb harm2_mag max v(rfo) [-1,2]
.meas hb fund1 param ='20*log10(fund1_mag)'
.meas hb fund2 param ='20*log10(fund2_mag)'
.meas hb harm1 param ='20*log10(harm1_mag)'
.meas hb harm2 param ='20*log10(harm2_mag)'
.meas hb IIP31 param ='fund1+(fund1-harm1)/2'
.meas hb IIP32 param ='fund1+(fund2-harm2)/2'
*

First, a voltage source element is used as a two-tone power source by setting

the power flag. A source impedance of 50 ohms is specified. The HB keyword is
used to identify the amplitude (interpreted as Watts with the power flag set to
1), phase, harmonic index, and tone index for each tone.
Vin rfind gnd dc=0 power=1 z0=50 $ 50 Ohm src
+ HB Pin 0 1 1 $ tone 1
+ HB Pin 0 1 2 $ tone 2

Second, the .HB command designates the frequencies of the two tones and
establishes the power sweep. The intmodmax parameter has been set to 7 to
include intermodulation harmonic content up to 7th order effects.
.HB tones=900MEG,910MEG nharms=11 intmodmax=7
+ SWEEP Pin dec 10 1e-8 1e-3

Last, the HSPICE advanced analog analyses ability to specify a specific

harmonic term is used in the .PRINT and .PROBE statements to pull out the
signals of particular interest. Notice the three different formats:
1. The following reference dumps a complete spectrum in RMS Watts for the
power across resistor Rload.
.PRINT HB P(Rload)

2. The following reference selectively dumps the power in resistor Rload at the
first harmonic of the 1st tone.
.PRINT HB P(Rload)[1,0]

3. The following reference selectively dumps the power in resistor Rload at the
3rd intermodulation product frequency (890 MHz).
.PRINT HB P(Rload)[2,-1]

HSPICE® User Guide: Advanced Analog Simulation and Analysis 25

N-2017.12
Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

To run this simulation, type the following at the command line:

hspice -i gsmlnaIP3_A.sp -o

Viewing Results Using Custom WaveView

For this analysis, the .PRINT statement will generate a
gsmlnaIP3_A.printhb0 file. Assume you want to find the output power
through the load resistor at the first tone, when the input power is 0.1mW.
To view the file:
1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the gsmlnaIP3_A.hb0 file.
3. Select the signal Pr(rload) [1,0] in the signal browser. Drag and drop
the signal in the waveform. The X-axis is the input power and the Y-axis is
the output power. The output power on the Y-axis is displayed in dBm but the
input power on the X-axis is displayed in watts. To make the 1dB
compression point measurement, it is necessary to change the X-axis scale
to dBm.
4. To change the X-axis scale to dBm, right-click the X-axis and select
Attributes from the menu. At the bottom right of the X-axis Attributes menu,
select dBm10 from the type pull-down menu and click Apply. Close the
menu by clicking Close.
5. To measure the 1dB compression point of the amplifier, select
Measurement from the tools menu. In the Measurement Tools window, the
available measurements are listed by category on the left side of the
window. Scroll down to the RF measurements and select P1dB. Click OK;
this places a dynamic meter in the waveform.
6. Move the dynamic meter near the signal, the meter will show the 1dB
compression point, the linear gain of the amplifier, and the input power
where the 1dB compression point is measured. For best results, the
asymptotic line drawn by the dynamic meter should overlay the linear
portion of the amplifier power curve.

26 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Tutorial Examples Harmonic Balance Analysis

Figure 3 1dB Compression Point

7. Use Waveview > New to open a new waveform.

8. Select the signals Pr(rload)[1,0], Pr(rload)[2,-1], and
Pr(rload)[2,0] in the signal browser. Drag and drop the signals in the
waveform. Change the X-axis scale to dBm as described in step 4.
9. The 3rd order intercept point is also measured by using the measurement
tool. In the measurement section, select IP3/SFDR. Click OK; this will place
a dynamic meter in the waveform.
10. Move the dynamic meter near the Pr(rload)[1,0] signal, the meter will
show the 3rd order intercept point and the input power where the 3rd order
intercept point is measured. For best results, the asymptotic lines drawn by
the dynamic meter should overlay the linear portion of the
Pr(rload)[1,0] and Pr(rload)[2,-1] signals.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 27

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Figure 4 3rd Order Intercept Point

Harmonic Balance Analysis

Harmonic balance analysis (HB) is a frequency-domain, steady-state analysis
technique. Use this analysis technique on a circuit excitable by DC and periodic
sources of one or more fundamental tones. The solution that HB finds is a set
of phasors for each harmonic signal in the circuit. You can think of this solution
as a set of truncated Fourier series. HSPICE advanced analog analyses allows
you to specify the solution spectrum to use in an analysis. HB analysis then
finds the set of phasors at these frequencies that describes the circuit
response. The result is a set of complex-valued Fourier series coefficients that
represent the waveforms at each node in the circuit.
Evaluation of linear circuit elements takes place in the frequency domain, while
evaluation of nonlinear elements occurs in the time domain. The nonlinear
response then transforms to the frequency domain where it adds to (or
“balances” with) the linear response. The resulting composite response

28 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

satisfies KCL and KVL (Kirchoff's current and voltage laws) with the circuit
solution.
Typical applications include performing intermodulation analysis, oscillator
analysis, and gain compression analysis, on amplifiers and mixers. HB analysis
also serves as a starting point for periodic AC and noise analyses.
For more information on control options, see .HB command in the HSPICE
Reference Manual: Commands and Control Options.
The following sections discuss these topics:
■
Features Supported
■
Harmonic Balance Equations
■
Steady-State HB Sources
■
Harmonic Balance Output Measurements
■
HB .PRINT and .PROBE Output Syntax
■
HB Output Data Files and Examples
■
HB Error and Warning Messages
■
Calculating Power Measurements After HB Analyses
■
Calculations for Time-Domain Output
■
Using .MEASURE with .HB Analyses

Features Supported
HB supports the following features:
■
All existing HSPICE advanced analog analyses models.
■
An unlimited number of independent input tones.
■
Sources with multiple HB specifications.
■
SIN, PULSE, VMRF, and PWL sources with TRANFORHB=1.
Prerequisites and Limitations
The following prerequisites and limitations apply to HB:
■
Requires one .HB statement.
■
Treats sources without a DC, HB, or TRANFORHB description as a
zero-value for HB, unless the sources have a transient description, in which
case, the time=0 value becomes a DC value.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 29

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Harmonic Balance Equations

We can write Kirchoff's current law in the time domain as:

Equation 1
t

f  v t  = i  v  t   + d q  v  t   +  y  t –  v    d + is  t  = 0
dt
–
■
i(v(t)) represents the resistive currents from nonlinear devices
■
q represents the charges from nonlinear devices
■
y represents the admittance of the linear devices in the circuit
■
is represents the vector of independent current sources
■
v is a variable that represents the circuit unknowns, both node voltages and
branch currents, and f(v,t) is an error term that goes to zero to satisfy
Kirchoff's current law.
Transforming this equation to the frequency domain results in:

Equation 2
F  V  = I  V  + Q  V  + Y   V + I s = 0

Note: Time-differentiation transforms to multiplication by j terms

(which make up the  matrix) in the frequency domain. The
convolution integral transforms to a simple multiplication. The Y
matrix is the circuit’s modified nodal admittance matrix.

All terms above are vectors that represent the circuit response at each analysis
frequency.
The following equation shows the vector of (complex-valued) unknowns in the
frequency domain for a circuit with K analysis frequencies and N unknowns.

Equation 3
V = V  1 0  V  1 1   V  1 K – 1  V  2 0   V  N K – 1 

HSPICE advanced analog analyses finds the unknown vector (V), to satisfy the
system of nonlinear equations, in the equation above. The Newton-Raphson
technique uses either a direct solver to factor the Jacobian matrix, or an indirect
solver. The HSPICE advanced analog analyses indirect solver is in the

30 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Generalized Minimum Residual (GMRES) Solver, a Krylov technique, and uses

a matrix-implicit algorithm.

Steady-State HB Sources
The fundamental frequencies used with harmonic balance analysis are
specified with the .HB TONES command. These frequencies can then be
referenced by their integer indices when specifying steady-state signal sources.
For example, the .HB specification given by the following line:
.HB TONES=1900MEG,1910MEG INTMODMAX=5
This specifies two fundamental frequencies: f  tone = 1  = 1.9GHz and
f  tone = 2  = 1.91GHz . Their mixing product at 10 MHz can then be referenced
using indices as f  2  – f  1  , while their 3rd order intermodulation product at 1.89
GHz can be referenced as 2f  1  – f  2  .
Steady-state voltage and current sources are identified with the HB keyword
according to
[HB [mag [phase [harm [tone [mod harm [modtone]]]]]]]
The source is mathematically equivalent to a cosine signal source that follows
the equation
A cos  t +  
where:
A = mag
 = 2 harm  f  tone  + modharm  f  mo dtone 

 = ---------  phase
180
Values for tone and modtone (an optional modulating tone) must be non-
negative integers that specify index values for the frequencies specified with
the .HB TONES command. Values for harm (harmonic) and modharm
(modulating tone harmonic) must be integers (negative values are valid) that
specify harmonic indices.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 31

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Examples

Example 3 The following example is a 1.0 Volt (peak) steady-state cosine voltage
source, which is at the fundamental HB frequency with zero phase and
with a zero volt DC value:
Vsrc in gnd DC 0 HB 1.0 0 1 1

Example 4 The following example is a steady-state cosine power source with 1.0mW
available power, which is implemented with a Norton equivalent circuit
and a 50 ohm input impedance:
Isrc in gnd HB 1.0e-3 0 1 1 power=1 z0=50

Example 5 Five series voltage sources sum to produce a stimulus of five equally
spaced frequencies at and above 2.44 GHz using modharm and
modtone parameters. These are commensurate tones (an integer
relation exists); therefore, you only need to specify two tones when
invoking the HB analysis.
.param Vin=1.0
.param f0=2440MEG
.param deltaf=312.5K
.param fcenter='f0 + 2.0*deltaf'
Vrfa in ina HB 'Vin' 0 1 1 $ 2.440625
GHz
Vrfb ina inb HB 'Vin' 0 1 1 -1 2 $
2.4403125 GHz
Vrfc inb inc HB 'Vin' 0 1 1 -2 2 $
2.440 GHz
Vrfd inc ind HB 'Vin' 0 1 1 +1 2 $
2.4409375 GHz
Vrfe ind gnd HB 'Vin' 0 1 1 +2 2 $ 2.44125
GHz
.HB tones=fcenter,deltaf intmodmax=5

Harmonic Balance Output Measurements

This section explains the harmonic balance output measurements you receive
after HSPICE runs an HB simulation.
The HB cosine sources interpret in real/imaginary and polar formats according
to Equation 4:

32 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Equation 4
j  t +   j jt
v  t  = A cos  t +   = Re  Ae  = Re  Ae e 
j
= Re  Ae  cos  t  + j sin  t   
= Re   V R + jV I   cos  t  + j sin  t   
= V R cos  t  – V I sin  at 
= A cos    cos  t  – A sin    sin  t 
This equation relates real/imaginary and polar formats with the standard
convention:

Equation 5
j
V R + jV I = Ae
V R = A cos   
V I = A sin   
2 2
A = VR + VI
VI
tan  = ------
VR
The result of HB analysis is a complex voltage (current) spectrum at each
circuit node (or specified branch). Let a[i] be the real part and b[i] be the
imaginary part of the complex voltage at the ith frequency index. The Fourier
series expansion gives the conversion to a steady-state time-domain waveform
as in Equation 6:

Equation 6
v(t) = a[0] + a[1]*cos(2f[1]*t) – b[1]*sin(2f[1]*t)
+ a[2]*cos(2f[2]*t) – b[2]*sin(2f[2]*t)
+ a[3]*cos(2f[3]*t) – b[3]*sin(2f[3]*t)
+...
+ a[N]*cos(2f[N]*t) – b[N]*sin(2f[N]*t)
Where:
■
v(t) is the resulting time domain waveform.
■
N+1 is the total number of harmonics (including DC) in the frequency
domain spectrum of the *.hb0 file (the zero-th data point represents DC).
■
a[i] is the real component at the ith frequency

HSPICE® User Guide: Advanced Analog Simulation and Analysis 33

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

■
b[i] is the imaginary component at the ith frequency
■
f[i] is the ith frequency value (with i=0 representing the zero frequency
DC term). These frequencies do not need a harmonic relationship.
This frequency domain (Fourier coefficient) representation converts to a
steady-state time domain waveform when you use the .PRINT or .PROBE
HBTRAN output option or you invoke the To Time Domain function on complex
spectra within Custom WaveView.

HB .PRINT and .PROBE Output Syntax

This section describes the syntax for the HB .PRINT and .PROBE statements.
.PRINT HB TYPE(NODES|ELEMENTS)[INDICES]

34 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

.PROBE HB TYPE(NODES|ELEMENTS)[INDICES]

Parameter Description

TYPE Specifies a harmonic type node or element.

TYPE can be one of the following:
■
Voltage type:
V = voltage magnitude and phase in degrees
VR = real component
VI = imaginary component
VM = magnitude
VP - Phase in degrees
VPD - Phase in degrees
VPR - Phase in radians
VDB - dB units
VDBM - dB relative to 1 mV
■
Current type:
I = current magnitude and phase in degrees
IR = real component
II = imaginary component
IM = magnitude
IP - Phase in degrees
IPD - Phase in degrees
IPR - Phase in radians
IDB - dB units
IDBM - dB relative to 1 mV
■ Power type: P = Power in Watts or Pdbm = Power in dBm
■
Frequency type:
‘HERTZ[i]’ (for single tone analysis), ‘HERTZ[i][j]’ (for two-tone analysis),
‘HERTZ[i][j][k]’ (for 3-tone analysis), etc.
You must specify the harmonic index integer for the HERTZ keyword. The
frequency of the specified harmonics results.

INDICES Index to tones in the form [n1, n2,..., nN], where nj is the index of the HB tone and the
HB statement contains N tones. Wildcards are illegal if you use the INDICES keyword.

TYPE can be one of the following:

■
Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■ Current type – an element name (elemname)
■
Power type – a resistor (resistorname) or port (portname) element name.

Use the following output syntax to transform HB data into the time domain:
.PRINT hbtran ov1 [ov2...]
.PROBE hbtran ov1 [ov2...]
Where ov1 ... are the output variables to print or probe.
Outputting Phase Noise Source as ASCII Data Files Using *.printpn0
HB phase noise and phase noise analyses can output simulation results as
ASCII data in *.printpn0 files for HBOSC and HBNOISE. Extend the E- and

HSPICE® User Guide: Advanced Analog Simulation and Analysis 35

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

G- voltage-controlled source syntax to make the phase noise data in ASCII

phase noise files double as input for specifying behavioral noise sources.
Usage Model
The syntax for the voltage controlled voltage (E) or current (G) source is as
follows:
Exxx node1 node2 noise file='filename' [mname='measname']
Gxxx node1 node2 noise file='filename' [mname='measname']
Where file='filename' is the name of the ASCII phase noise data file. The
file 'design.printpn0', name is typically designates an .HBOSC phase noise
analysis or .HBNOISE analysis output file.
Use mname='measname' to select the appropriate noise measurement name
from the *.printpn0 file.
measname can be one of the following:
■
NLP_L(f) - selects the nlp_L(f) phase noise data in units of dBc/Hz
■
PAC_L(f) - selects the pac_l(f) phase noise data in units of dBc/Hz
■
BPN_L(f) - selects the bpn_l(f) phase noise data in units of dBc/Hz
■
ONOISE - selects the onoise data based on .HBNOISE or .SNNOISE
analysis

HB Output Data Files and Examples

The results of an HB analysis are complex spectral components at each
frequency point. The a[i] is the real part, and b[i] is the imaginary part of the
complex voltage at frequency index i. The conversion to a steady-state time-
domain is then given by the Fourier series expansion.
An HB analysis produces these output data files:
■ Output from the .PRINT HB statement is written to a .printhb# file.
• The header contains the large signal fundamental frequencies.
• The columns of data are labeled as HERTZ, followed by frequency
indices, and then the output variable names.
• The sum of the frequency indices, multiplied by the corresponding
fundamental frequencies, add up to the frequency in the first column.

36 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

■
Output from the .PROBE HB statement is written to a .hb# file. It is in the
same format as the HSPICE transient analysis .tr# file. Besides the output
waveform, it contains the information of harmonic indices and basic tone
frequencies.
■
Output from the .PRINT HBTRAN statement is written to a printer file. The
format is identical to a .print# file.
■
Output from the .PROBE HBTRAN statement is written to a .hr# file. The
format is identical to a .tr# file.
■
Reported performance log statistics are written to a .lis file.
.HB Output Example
.PRINT HB P(rload) $ RMS power (spectrum)
$ dissipated at the rload resistor
.PROBE HB V(n1,v2) $ Differential voltage (spectrum)
$ between the n1,n2 nodes
.PRINT HB VP(out)[1] $ Phase of voltage at the out
$ node, at the fundamental
$ frequency
.PROBE HB P(Pout)[2,-1] $ RMS power delivered to the Pout
$ port, at third-order intermod
.PRINT HBTRAN V(n1) $ Voltage at n1 in time domain
.PROBE HBTRAN V(n1, n2) $ Differential voltages between n1
$ and n2 node in time domain.

HB Error and Warning Messages

The following sections discuss these topics:
■ HB Analysis Error Messages
■
HB Analysis Warning Messages

HB Analysis Error Messages

Following are the error messages issued by HSPICE simulator when using the
.HB analysis:

File Description

HB_ERR.1 Harmonic numbers must be positive non-zero.

HB_ERR.2 No .hb frequencies given.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 37

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

File Description

HB_ERR.3 Negative frequency given.

HB_ERR.4 Number of harmonics should be greater than zero.

HB_ERR.5 Different number of tones, nharms.

HB_ERR.6 Bad probe node format for oscillator analysis.

HB_ERR.7 Bad format for FSPTS.

HB_ERR.8 Bad .hb keyword.

HB_ERR.9 Tones must be specified for .hb analysis.

HB_ERR.10 Nharms or intmodmax must be specified for .hb analysis.

HB_ERR.11 Source harmonic out of range.

HB_ERR.12 Source named in the tones list is not defined.

HB_ERR.13 Source named in the tones list does not have TRANFORHB specified.

HB_ERR.14 Source named in the tones list has no transient description.

HB_ERR.15 Source named in the tones list must be HB, SIN, PULSE, PWL, or VMRF.

HB_ERR.16 Tone specification for the source is inconsistent with its frequency.

HB_ERR.17 HB oscillator analysis has reached the NULL solution.

HB_ERR.18 Bad subharms format.

HB_ERR.19 Modtone may not be set to the same value as tone.

HB Analysis Warning Messages

Following are the warning messages issued by HSPICE when using the .HB
analysis:

File Description

HB_WARN.1 .hb multiply defined. Last one will be used.

HB_WARN.2 Tone specified for V/I source not specified in .HB command.

38 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

File Description

HB_WARN.3 HB convergence not achieved.

HB_WARN.4 Source specifies both HB and transient description. HB description will be

used.

HB_WARN.5 Source specifies exponential decay. HB will ignore it.

HB_WARN.6 Source specifies a non-positive frequency.

HB_WARN.7 Source does not fit the HB spectrum.

HB_WARN.8 Source cannot be used with the TRANFORHB option.

HB_WARN.9 Frequency not found from transient analysis

HB_WARN.10 .hb/.hbosc will be ignored due to .env/.envosc.

Calculating Power Measurements After HB Analyses

Two types of power measurements are available: dissipated power in resistors
and delivered power to port elements.
The following sections discuss these topics:
■
Power Dissipated in a Resistor
■
Power Delivered to a Port Element
■
Examples

Power Dissipated in a Resistor

All power calculations make use of the fundamental phasor power relationship
given in the following equation, where voltage V and current I are complex
phasors given in peak values (not rms nor peak-to-peak):

Equation 7
1
P rms = --- Re  VI 
2
In the case of a simple resistor, its current and voltage relate to each other
according to Vn=InR. Equation 8 gives the power dissipated in a resistor of
(real) value R at frequency index n :

HSPICE® User Guide: Advanced Analog Simulation and Analysis 39

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Equation 8
2
Vn
P rms  resistor   n  = -----------
2R

Power Delivered to a Port Element

The port element can be either a source or sink for power. You can use a
special calculation that computes the power flowing into a port element even if
the port element itself is the source of that power. In Figure 5 a Port element
connects to a circuit (the Port element may or may not include a voltage
source).

Zo In

+
+
Remainder
Port
Vn of
Vs Element
Circuit
-

Figure 5 Port Element

Let Vn be the (peak) voltage across the terminals of the port element (at
frequency index n). Let In be the (peak) current into the (1st) terminal of the
port element (at frequency index n). Let Zo be the impedance value of the z0
port element. Then, you can compute the power wave flowing into the terminals
of the port element (at frequency index n) according to:

Equation 9
1 Vn + Zo In 2
P in  n  = --- ----------------------
-
2 2 Z
o
This power expression remains valid whether or not the port element includes
an internal voltage source at the same frequency. If the port element includes a

40 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

voltage source at the same frequency, you can use this power calculation to
compute the magnitude of the related large-signal scattering parameters.
If you expand the preceding formula, you can determine the power delivered to
a port element with (real) impedance Zo:

Equation 10
2 2 2
1  Vn + Zo In 1 
P rms  port   n  = ---  ------------------------------------ + --- Re  V n In  
2 4Z o 2 
This power value represents the power incident upon and delivered to the port
element's load impedance (Zo) due to other power sources in the circuit, and
due to reflections of its own generated power.
If you use the port element as a load resistor (no internal source), the
preceding equation reduces to that for the simple resistor.
If you use the port element as a power source (with non-zero available power,
that is, a non-zero Vs) and it terminates in a matched load (Zo), the port
power-measurement returns 0 W, because no power is reflected.
You can request power measurements in the form of complete spectra or in the
form of scalar quantities that represent power at a particular element. To
request a complete power spectrum, use the following syntax.
.PRINT HB P(Elem)
.PROBE HB P(Elem)
To request a power value at a particular frequency tone, use the following
syntax:
.PRINT HB P(Elem)[<n1<,n2<n3,...>>>]
.PROBE HB P(Elem)[<n1<,n2<,n3,...>>>]
The Elem is the name of either a Resistor (R) or Port (P) element, and n1, n2,
and n3 are integer indices used for selecting a particular frequency in the
Harmonic Balance output spectrum.

Examples

Example 6 Prints a table of the RMS power (spectrum) dissipated by resistor R1.
.PRINT HB P(R1)

HSPICE® User Guide: Advanced Analog Simulation and Analysis 41

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Example 7 Outputs the RMS power dissipated by resistor R1 at the fundamental HB

analysis frequency following a one-tone analysis.
.PROBE HB P(R1)[1]

Example 8 Prints the power dissipated by resistor R1 at DC following a one-tone

analysis.
.PRINT HB P(R1)[0]

Example 9 Outputs the RMS power dissipated by resistor R1 at the (low-side) 3rd
order intermodulation product after an HB two-tone analysis.
.PROBE HB P(R1)[2,-1]

Example 10 Prints the RMS power dissipated by resistor R1 at the (high-side) 3rd
order intermodulation product after an HB two-tone analysis.
.PRINT HB P(R1)[-1,2]

Example 11 Outputs the RMS power (spectrum) delivered to port element Pload.
.PROBE HB P(Pload)

Example 12 Prints the RMS power delivered to port element Pload at the fundamental
HB analysis frequency following a one-tone analysis.
.PRINT HB P(Pload)[1] $

Example 13 Outputs the RMS power delivered to port element Pload at the (low-side)
3rd order intermodulation product after an HB two-tone analysis.
.PROBE HB P(Pload)[2,-1]

Calculations for Time-Domain Output

In addition to a frequency-domain output, HB analysis also supports a time-
domain output. The simulation generates an equivalent time-domain waveform
according to the Fourier series expansion by way of

Equation 11

V  n1 @time t = SUM OVERm  REALV  n1   m    cos ""    m   t  – IMAG(V  n1   m  sin   m  t t

Where m starts from 0 to the number of frequency points in the HB simulation.

The output syntax is
.PRINT [HBTRAN | HBTR] V(n1)
.PROBE [HBTRAN | HBTR] V(n1)

42 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

The output time ranges from 0 to twice the period of the smallest frequency in
the HB spectra.
Minimizing Gibbs Phenomenon
You can use the HB_GIBBS option for HBTRAN output to minimize Gibbs’
phenomenon that may occur in transforming a square-wave signal from the
frequency domain to the time domain. The syntax is .OPTION HB_GIBBS=n
(defaults to zero, which is equivalent to not using it at all). The result is that the
N
A  sin c  x   function filters HBTRAN waveforms before they transform to the
time domain via FFT. This option applies only to single-tone output. For
example:
.option hb_gibbs = 2
...
.print hbtran v(2)

Figure 6 Upper square-wave signal shows HB_GIBBS = 2, while the lower shows
the option = 0

HSPICE® User Guide: Advanced Analog Simulation and Analysis 43

N-2017.12
Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Using .MEASURE with .HB Analyses

■
For transient analysis (TRAN), the independent variable for
calculating .MEASURE is time.
■
For AC analysis, the independent variable for calculating .MEASURE is
frequency.
■
However, as with DC analysis, the use of a .MEASURE command is peculiar
for HB analysis, because it has no obvious independent variable.
In HSPICE advanced analog analyses, the independent variable for HB
.MEASURE analysis is the first swept variable specified in the .HB simulation
control statement. This variable can be anything: frequency, power, voltage,
current, a component value, and so on.
If you do not specify an HB sweep, then .MEASURE assumes a single-valued
independent variable sweep.
You can apply the measurements to current, voltage, and power waveforms.
The independent variable for measurements is the swept variable (such as
power), not the frequency axis that corresponds to a single HB steady state
point.
HSPICE advanced analog analyses also support the .MEASURE [HBTRAN |
HBTR] ... syntax. Similar to the .PROBE and .PRINT HBTR statements in
the section Calculations for Time-Domain Output, you apply a .MEASURE HBTR
statement on the signals obtained in the same way. Moreover, like a .MEASURE
statement in transient analysis, the independent variable in a .MEASURE HBTR
statement is time.
HSPICE advanced analog analyses optimization can read the data from
.MEASURE HB and .MEASURE HBTR statements. Due to the difference in the
independent variable between the .MEASURE HB and .MEASURE HBTR
statements, you cannot mix these two types of measurements in an HSPICE
advanced analog analyses optimization. But you can combine a .MEASURE
HBTR statement with a .MEASURE PHASENOISE statement (see Measuring
Phase Noise with .MEASURE PHASENOISE) and a .MEASURE HBNOISE
statement (see Measuring HBNOISE Analyses with .MEASURE) in an HSPICE
advanced analog analyses optimization flow.

44 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Harmonic Balance Analysis

Examples

Example 14 For the following .HB simulation control statement, the independent
variable is the swept tone frequency, and the .MEASURE command
values return results based on this frequency sweep:
* HARMONIC BALANCE tone-frequency sweep for amplifier
.param freq1=1.91e9 power=1e-3
.HB tones=freq1 nharms=10 sweep freq1 LIN 10 1.91e9 2.0e9
.MEASURE HB Patf0 FIND P(Rload)[1] AT=1.95e9 $ Power at
+ f0=1.95Ghz
.MEASURE HB Frq1W WHEN P(Rload)[1]=1. $ freq1 @ 1 Watt
.MEASURE HB BW1W TRIG AT=1.92e9 TARG P(Rload)[1] VAL=1.
+ CROSS=2 $ 1 Watt bandwidth
.MEASURE HB MaxPwr MAX P(Rload)[1] FROM=1.91e9 TO=2.0e9
+ $ Finds max output power
.MEASURE HB MinPwr MIN P(Rload)[1] FROM=1.91e9 TO=2.0e9
+ $ Finds min output power

Example 15 In the following example, the independent variable is the power variable,
and the .MEASURE values return results based on the power sweep.
Units are in Watts.
* HARMONIC BALANCE power sweep for amplifier
.param freq1=1.91e9 power=1e-3
.HB tones=freq1 nharms=10 sweep power DEC 10 1e-6 1e-3
.MEASURE HB Pat1uW FIND P(Rload)[1] AT=1e-6 $ Pout at 1uW
.MEASURE HB Pin1W WHEN P(Rload)[1]=1. $ Pin @ 1 Watt Pout
.MEASURE HB Prange1W TRIG AT=1.92e9 TARG P(Rload)[1] VAL=1.
+ CROSS=2 $ 1W oper. range

.MEASURE HB ssGain DERIV P(Rload)[1] AT=1e-5

+ $ relative power gain at 10uW input
.MEASURE HB Gain3rd DERIV P(Rload)[3] AT=1e-5
+ $ 3rd harmonic gain at 10uW input
.MEASURE HB PAE1W FIND ‘(P(Rload)[1]-power)/P(Vdc)[0]’
+ WHEN P(Rload)[1]=1 $ PAE at 1 Watt output

HSPICE® User Guide: Advanced Analog Simulation and Analysis 45

N-2017.12
Chapter 2: Harmonic Balance Analysis
Troubleshooting Convergence Issues with Harmonic Balance Analysis

Example 16 In this example, the independent variable is again the power variable, and
the .MEASURE values return results based on the power sweep. This is a
two-tone sweep, where both input frequency sources are at the same
power level in Watts.
* HARMONIC BALANCE two-tone sweep for amplifier
* An IP3 calculation is made at 10uW in the sweep
.param freq1=1.91e9 freq2=1.91e9 power=1e-3
.HB tones=freq1,freq2 nharms=6,6 sweep power DEC 10 1e-6 1e-3
.MEASURE HB Pf1dBm FIND ’10.*LOG(P(Rload)[1,0]/1.e-3)’
+ AT=1e-5 $ P(f1) at 10uW input
.MEASURE HB P2f1_f2dBm FIND ’10.*LOG(P(Rload)[2,-1]/1.e-3)’
+ AT=1e-5 $ P(2f1-f2) at 10uW input
.MEASURE HB OIP3dBm PARAM = ‘0.5*(3.*Pf1dBm-P2f1_f2dBm)’
.MEASURE HB IIP3dBm PARAM = ‘OIP3dBm-Pf1dBm+20.0’
.MEASURE HB AM2PM DERIV VP(outp,outn)[1] AT=1e-5
+ $ AM to PM Conversion in Deg/Watt

Troubleshooting Convergence Issues with Harmonic

Balance Analysis
To troubleshoot convergence issues with harmonic balance analysis, perform
the following steps:
■
Ensure that the transient sources are activated by setting .option
tranforhb.
■
The number of harmonics (nharms) specified in the .HB command will
affect the accuracy. The number of harmonics should be set to
approximately 40 per the fastest period in the circuit. For example, for a
circuit containing a divide by 6, at least 240 harmonics should be used.
■ To verify that the correct number of harmonics is used, observe the time
domain waveforms. To output the time domain waveforms, include .PROBE
hbtr statements in the netlist. For example, .probe hbtr v(*). The time
domain waveform file will have the extension hr0.
■
Increase the initial transient analysis time by increasing the setting of the
option HBTRANINIT to a value equal to 100 times the waveform period.
■
Increase the value of the number of harmonics (nharms) in the .HB
command.

46 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Phase Differences Between HB and SIN Sources

Phase Differences Between HB and SIN Sources

The HB steady-state cosine source has a phase variation compared to the
TRAN time-domain SIN source. The SIN source (with no offset, delay or
damping) follows the equation:

Equation 12
A sin  t +  
while the HB sources follow:

Equation 13
A cos  t +  
In order for the two sources to yield identical results it is necessary to align
them by setting their phase values accordingly using:

Equation 14
A cos  t +   = A sin  t +  + 90 

Equation 15
A sin  t +   = A cos  t +  – 90 

Example
Example 17 To specify sources with matching phase for HB and TRAN analysis, use
a convention similar to:
** Example #1 with equivalent HB and SIN sources
** SIN source is given +90 phase shift
.param freq1=2400MEG Vin=1.0
Vsrc in gnd DC 0 HB 'Vin' 0 1 1 SIN(0 'Vin' 'freq1' 0 0 90)
.HB tones=freq1 intmodmax=7
** Example #2 with equivalent HB and SIN sources
** HB source is given -90 phase shift to align with SIN
.param freq1=2400MEG Vin=1.0
Vsrc in gnd DC 0 HB 'Vin' -90 1 1 SIN(0 'Vin' 'freq1' 0)
.HB tones=freq1 intmodmax=7
** Example #3 with equivalent .HB and .TRAN sources
** SIN source is activated for HB using "TRANFORHB"
.param freq1=2400MEG Vin=1.0
Vsrc in gnd DC 0 SIN(0 'Vin' 'freq1' 0) TRANFORHB=1
.HB tones=freq1 intmodmax=7

HSPICE® User Guide: Advanced Analog Simulation and Analysis 47

N-2017.12
Chapter 2: Harmonic Balance Analysis
Optimization

Optimization
To perform optimization, create an input netlist file that specifies:
■
Optimization parameters with upper and lower boundary values along with
an initial guess.
■
An HB or HBOSC optimization statement.
■
An optimization model statement.
■
Optimization measurement statements for optimization parameters.
If you provide the input netlist file, optimization specifications, limits, and initial
guess, then the optimizer reiterates the simulation until it finds an optimized
solution.
Usage Notes and Examples
■
Optimization works for HB, HBOSC, and HBAC analyses.
■
You can add the GOAL options in every meaningful .MEASURE statement,
like FIND-WHEN, FIND-AT, and so on.
■
A data sweep does not need definition in the .HB statement for HB
optimization to use the measured result from .MEASURE HBNOISE, or
PHASENOISE statements. Therefore, this type of optimization does not
support parameter sweep.
■
Optimize multiple parameters with multiple goals by selecting .MODEL OPT
LEVEL=0 (modified Lavenberg-Marquardt method).
■
Optimize single parameters in single measurement situations by selecting
.MODEL OPT LEVEL=1 (bisection method).
■ Examples
• Setting optimization parameters
.param W=opt1(231u, 100u, 800u)
.param Rs=opt1(10,8,20)

• Optimization analysis statement

.HB tones=2.25g 2.5g nharms=6,3
+ sweep Pin_dbm -30 0 2
+ sweep optimize = opt1
+ results = gain $measure result to tune the parameters
+ model= optmod1

• Selecting an optimization model

48 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 2: Harmonic Balance Analysis
Optimizing with HB Measurements

.model optmod1 opt level=1 $Bisection method

+ itropt=40 relin=1e-4 relout=1e-6 $ accuracy settings

• Measurement statements to tune the optimization parameters

.measure HB vif find vdb(if+)[-1,1] at 10e-6
.measure HB vrf find vdb(rf+)[0,1] at 10e-6
.measure HB gain=param('vif-vrf') goal=-2

• Measurement statement to find the fundamental frequency from HB

analysis:
.measure HB frequency_max FIND ‘HERTZ[1]’ at=0

Optimizing with HB Measurements

The required statements are:
■
Analysis statement
.HB TONES=f1[f2 ... fn] [NHARMS=h1 [,h2 ... hn]]
+ SWEEP parameter_sweep OPTIMIZE=OPTxxx RESULT=measname
+ MODEL=mname
■
Measure statement
.MEASURE HB measname FIND out_var1 AT=val GOAL=val

HSPICE® User Guide: Advanced Analog Simulation and Analysis 49

N-2017.12
Chapter 2: Harmonic Balance Analysis
References

References
[1] S. Maas, Nonlinear Microwave Circuits, Chapter 3, IEEE Press, 1997.
[2] R. Gilmore and M.B. Steer, “Nonlinear Circuit Analysis Using the Method of
Harmonic Balance - A Review of the Art, Part I, Introductory Concepts.”
International Journal of Microwave and Millimeter-wave Computer-Aided
Engineering, Volume 1, No. 1, pages 22-37, 1991.
[3] R. Gilmore and M.B. Steer, “Nonlinear Circuit Analysis Using the Method of
Harmonic Balance - A Review of the Art. Part II. Advanced Concepts.”
International Journal of Microwave and Millimeter-wave Computer-Aided
Engineering, Volume 1, No. 2, pages 159-180, 1991.
[4] V. Rizzoli, F. Mastri, F. Sgallari, G. Spaletta, “Harmonic-Balance Simulation
of Strongly Nonlinear Very Large-Size Microwave Circuits by Inexact
Newton Methods,” MTT-S Digest, pages 1357-1360, 1996.
[5] S. Skaggs, Efficient Harmonic Balance Modeling of Large Microwave
Circuits, Ph.D. thesis, North Carolina State University, 1999.
[6] R.S. Carson, High-Frequency Amplifiers, 2nd Edition, John Wiley & Sons,
1982
[7] S.Y. Liao, Microwave Circuit Analysis and Amplifier Design, Prentice-Hall,
1987.
[8] J. Roychowdhury, D. Long, P. Feldmann, “Cyclostationary Noise Analysis of
Large RF Circuits with Multitone Excitations”, IEEE JSCC, volume 33,
number 3, March 1998.
[9] Y. Saad, Iterative Methods for Sparse Linear Systems, PWS Publishing
Company, 1995.
[10] J. Roychowdhury, D. Long, and P. Feldmann, “Cyclostationary Noise
Analysis of Large RF Circuits with Multitone Excitations,” IEEE Journal of
Solid-State Circuits, volume 33, pages 324–336, March 1998.
[11] K. Kurakawa, “Power waves and the Scattering Matrix,” IEEE Trans.
Microwave Theory Tech., vol. MTT-13, pp. 194-202, March 1965.

50 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 3
3 Shooting Newton Analysis

Describes HSPICE advanced analog steady-state time domain analysis based

on an advanced Shooting-Newton algorithm.

The following sections discuss these topics:

■
Shooting Newton Steady-State Time Domain Analysis (.SN)
■ Shooting Newton with Fourier Transform (.SNFT)
■
Troubleshooting Convergence Issues with Shooting Newton Analysis

Shooting Newton Steady-State Time Domain Analysis

(.SN)
An advanced Shooting Newton (SN) algorithm provides additional performance
and functionality to HSPICE advanced analog analyses for time-domain,
steady-state analysis.
Shooting-Newton adds analysis capabilities for PLL components, digital
circuits/logic such as ring oscillators, frequency dividers, phase/frequency
detectors (PFDs), switched capacitor filters, and for other digital logic circuits
and RF components that require steady-state analysis, but operate with
waveforms that are more square wave than sinusoidal.
For more information on control options, see .SN command in the HSPICE
Reference Manual: Commands and Control Options.
The Shooting-Newton algorithm effectively analyzes applications including:
■
Ring oscillators (see Chapter 4, Oscillator Analysis)
■
Frequency dividers (prescalers)

HSPICE® User Guide: Advanced Analog Simulation and Analysis 51

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

■
Mixer conversion gain
■
Phase-frequency detectors (PFDs)
■
Mixer noise figure
Functionality includes:
■
Both driven and oscillator (autonomous) analyses
■
Time domain or frequency analysis based on advanced Shooting Newton
algorithm
■
Shooting Newton with Fourier Transform (.SNFT)
■
Shooting Newton AC Analysis (.SNAC)
■
Shooting Newton Oscillator Analysis (.SNOSC)
■
Oscillator Analysis
This section covers the following topics:
■
Shooting Newton Analysis — Tutorial Example
■
SN .PRINT and .PROBE Output Syntax

Shooting Newton Analysis — Tutorial Example

While the Harmonic Balance (HB) algorithm represents the circuit's voltage and
current waveforms as a Fourier series (a series of sinusoidal waveforms), the
Shooting Newton (SN) algorithm provides analysis capability for digital logic
circuits and advanced analog components that require steady-state analysis,
but operate with waveforms that tend to be square instead of sinusoidal.
This section covers the following topics:
■
Shooting Newton Analysis — Setup
■
Clock Buffer Analysis Example

52 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

Shooting Newton Analysis — Setup

To set up a time-domain, steady-state analysis, the HSPICE input netlist must
contain:
■
A .SN command to activate the analysis. The .SN command specifies:
■
The expected period of the steady-state waveforms, which must match the
period of any input waveforms. The period is specified in time domain units
(seconds). Alternatively, you may specify a frequency in Hz.
■
A time resolution, which is analogous to the transient analysis (.TRAN)
command’s TSTEP parameter and will affect the time-step size selection. It
also affects the number of frequency terms used in small-signal analyses,
such as periodic AC or noise analysis. The time resolution is typically
specified in seconds but, alternatively, may be specified in the frequency
domain as a number of harmonics.
■
A transient initialization time that is used by HSPICE advanced analog
analyses to run a basic transient simulation of this length before attempting
Newton-Raphson iterations to converge on a steady-state solution. This
parameter is optional. If it is not specified, the specified period is used as the
initialization time. The initial transient analysis is used for circuit stabilization
before the steady state solution is found. Larger initialization values typically
result in convergence that is more robust.
■
For oscillator circuits, a .SNOSC command is used to activate the analysis.
The .SNOSC command specifies:
■
The approximate frequency of oscillation specified either as a frequency (in
Hertz) or as the time domain period.
■ The number of high frequency harmonics. Alternatively, a time resolution in
seconds can be specified.
■
A transient initialization time that is used by HSPICE advanced analog
analyses to run a basic transient simulation of this length before attempting
Newton-Raphson iterations to converge on a steady state solution. This
parameter is optional. If it is not specified, the period of the specified

HSPICE® User Guide: Advanced Analog Simulation and Analysis 53

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

frequency of oscillation is used as the initialization time. For oscillators we

recommend specifying a transient initialization time since the default
initialization time is usually too short to effectively stabilize the circuit.
■
A node at which to probe for oscillation conditions.
■
If the tuning curve of a VCO is to be analyzed, the optional parameter
MAXTRINITCYCLES can be specified.
■
One or more signal sources for driving the circuit in SN analysis, if the circuit
is driven. In the case of autonomous oscillator analysis, no signal source is
required. The sources are required to be time domain sources and must
match the period specified in the .SN command.

Clock Buffer Analysis Example

This example demonstrates the Shooting Newton-based analysis of a clock
buffer circuit. Extracted portions of the input file are presented in the following
pages. The complete clockbuf4sn.sp input file for this example is located at
$installdir/demo/hspice/rf_examples/.

Example 18 Clock Buffer Netlist

**
** Shooting Newton analysis example
**
.temp 30.0
.option scale=50n
.option post probe
.subckt inverter VDD in out
M1 out in 0 0 nmos L=1 W=10
M2 out in VDD VDD pmos L=1 W=20
.ends
*-----------------------------------------------------------
* reference clock
*
.param vdd=0.80
.param freq=2000MEG
.param period='1.0/freq'
Vsrc ref gnd DC 0
+ sin (0 'vdd' 'freq')

54 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

*------------------------------------------------------------
* Circuit:
*
Vdd VDD 0 DC 'vdd'
Xrout VDD ref outa inverter
* Chain of 4 inverters
Xb11 VDD outa outb11 inverter
Xb12 VDD outb11 outb12 inverter
Xb13 VDD outb12 outb13 inverter
Xb14 VDD outb13 outb14 inverter
RLb1 outb14 0 100k
* Chain of 6 inverters
Xb21 VDD outa outb21 inverter
Xb22 VDD outb21 outb22 inverter
Xb23 VDD outb22 outb23 inverter
Xb24 VDD outb23 outb24 inverter
Xb25 VDD outb24 outb25 inverter
Xb26 VDD outb25 outb26 inverter
RLb2 outb26 0 100k
*------------------------------------------------------------
* SN analysis setup
*
*
* Using frequency domain syntax
.sn tone='freq' nharms=63 trinit='10*period' numperout=2
*
* Using time domain syntax
.sn tres='0.02*period' period='period' trinit='10*period'
+ numperout=2
*
.snnoise v(outb26) Vsrc dec 10 100 100meg [0,1]
*
*
.probe sn v(ref) v(outa) v(outb14) v(outb26)
.probe snfd v(ref) v(outb26)
.probe snnoise onoise
*

The .SN command can be specified using either the frequency domain syntax
that specifies the tone and number of harmonics, or the time domain syntax
that specifies the period and time resolution. In both cases, a transient
initialization time is set and the number of output periods of the signal are
specified using the NUMPEROUT keyword.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 55

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

* Using frequency domain syntax

.sn tone='freq' nharms=63 trinit='10*period' numperout=2

* Using time domain syntax

.sn tres='0.05*period' period='period' trinit='10*period'
numperout=2

The steady state noise analysis

.snnoise v(outb26) Vsrc dec 10 100 100meg [0,1]

A periodic, time-varying AC noise analysis based on the Shooting Newton

algorithm is performed using the .SNNOISE command. The .SNNOISE
analysis requires an output node (v(outb26)) where the noise is to be
measured, an input noise source (Vsrc) which serves as the reference for the
noise computation, and a frequency sweep for the noise analysis. Optionally,
an index term can be defined. The index term specifies the output frequency
band at which the noise is evaluated. For this case, you will evaluate the low
frequency noise of the clock buffer.
During the simulation, the simulation status is displayed on the computer
screen. In addition to the screen display, a more detailed status, cpu time, and
memory usage information are also written to the clockbuf4sn.lis file.
To run this simulation, type the following at the command line:
hspice -i clockbuf4sn.sp -o

Viewing Results Using Custom WaveView

You can view the steady-state time domain, frequency domain and noise
results.
To view the time domain results:
1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the clockbuf4sn.sn0 file.
3. Select the signals v(outb26) and v(outb14) in the signal browser. Drag
and drop the signals in the waveform.

56 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

Figure 7 SN Analysis Output: Time Domain

To view the frequency domain results:

1. Use File > Import Waveform File and select the clockbuf4sn.snfd0
file.
2. Use Waveview > New to open a new waveform.
3. Select the signal v(outb26) in the signal browser. Drag and drop the
signal in the waveform.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 57

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

Figure 8 SN Analysis Output: Frequency Domain

To view the noise analysis results:

1. Use File > Import Waveform File and select the clockbuf4sn.snpn0
file.
2. Use Waveview > New to open a new waveform.
3. Select the signal onoise() in the signal browser. Drag and drop the signal in
the waveform.
4. The X-axis scale will be linear and it is preferred to use the log scale.
To change the X-axis scale to the log scale, right-click the X-axis. A menu
will appear. Select log in this menu.

58 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

Figure 9 SNNOISE Analysis Results

SN .PRINT and .PROBE Output Syntax

The output from .SN analysis is generated in both time and frequency domains.
The time domain output variables are the same as for standard transient
analysis:
■ individual nodal voltages: V(n1 [,n2])
■
branch currents: I(Vxx)
■
element power dissipation: In(element)
It is also possible to output the results from Shooting Newton analysis in terms
of complex, frequency-domain output variables. This output format is activated
by using the “SNFD” keyword in the output syntax.
For output in the frequency domain, the syntax is identical to the Harmonic
Balance output syntax:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 59

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton Steady-State Time Domain Analysis (.SN)

.PRINT SNFD TYPE (NODES|ELEMENTS)[INDICES]

.PROBE SNFD TYPE (NODES|ELEMENTS)[INDICES]

Parameter Description

SNFD TYPE Specifies a harmonic type node or element.

TYPE can be one of the following:
■
Voltage type:
V = voltage magnitude and phase in degrees
VR = real component
VI = imaginary component
VM = magnitude
VP - Phase in degrees
VPD - Phase in degrees
VPR - Phase in radians
VDB - dB units
VDBM - dB relative to 1 mV
■ Current type:
I = current magnitude and phase in degrees
IR = real component
II = imaginary component
IM = magnitude
IP - Phase in degrees
IPD - Phase in degrees
IPR - Phase in radians
IDB - dB units
IDBM - dB relative to 1 mV
■
Power type – P
■ Frequency type: hertz[index], hertz[index1, index2, ...]. You must specify the
harmonic index for the hertz variable. The frequency of the specified harmonics is
dumped.

INDICES Is the harmonic index of the SNFD tone. Index is limited to the single tone associated
with the SN analysis.

SNFD TYPE can be one of the following:

■
Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■
Current type – an element name (elemname)
■
Power type – a resistor (resistorname) or port (portname) element name.

Output Files
The time domain data are output to printsn0 and .sn0 files. Frequency
domain data are output to .printsnf0 and .snf0 files.
Output Format
The format for time domain output is the same as standard transient analysis.
For frequency domain output, the format is similar to HB. The main difference is
that the Shooting Newton output in the frequency domain is single tone only.

60 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Shooting Newton with Fourier Transform (.SNFT)

The results of an SN analysis are complex spectral components at each

frequency point. The a[i] is the real part and b[i] is the imaginary part of the
complex voltage at frequency index i. The conversion to a steady-state time-
domain is then given by the Fourier series expansion.
An SN analysis produces these output data files:
■
Output from the .PRINT SN statement is written to a .printsn# file.
• The header contains the large signal fundamental frequencies.
• The columns of data are labeled as HERTZ, followed by frequency
indices, and then the output variable names.
• The sum of the frequency indices, multiplied by the corresponding
fundamental frequencies, add up to the frequency in the first column.
■
Output from the .PROBE SN statement is written to a .sn# file in the same
format as the HSPICE transient analysis .tr# file. It contains the
information on harmonic indices and basic tone frequencies plus the output
waveform.
■
Reported performance log statistics are written to a .lis file:
• Name of SN data file.
• Simulation time:
DC operating point (op) time
SN time
Total simulation time
• Memory used
• Size of matrix (nodes * harmonics)
• Final SN residual error

Shooting Newton with Fourier Transform (.SNFT)

The .SNFT command is to the .SN analysis what .FFT is to the TRAN
analysis, a means to provide spectrum analysis. Spectrum analysis represents
a time-domain signal, within the frequency domain. .SNFT uses the Fourier
transform: a Discrete Fourier Transform (DFT) uses sequences of time values
to determine the frequency content of analog signals, in circuit simulation.
The .SNFT statement uses the internal time point values.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 61

N-2017.12
Chapter 3: Shooting Newton Analysis
Shooting Newton with Fourier Transform (.SNFT)

By default, the .SNFT statement uses a second-order interpolation to obtain

waveform samples, based on the number of points that you specify.

Note: True distributed components (such as ideal delays or

transmission lines) are not supported; components with hidden
states are not supported.

SN analysis assumes that all stimuli are periodic with period T. If the circuit is
driven with more than one periodic stimulus, then the frequencies must be all
co-periodic and T must match the common period or some integer multiple of it.
The .SN analysis only supports .tran (time-domain) periodic signal sources.
(Refer to the .tran analysis for a detailed documentation on transient signal
sources).
You can use windowing functions to reduce the effects of waveform truncation
on the spectral content. You can also use the .SNFT command to specify:
■
output format
■
frequency
■
number of harmonics
■
total harmonic distortion (THD)
For more information, see the .SNFT command in the HSPICE Reference
Manual: Commands and Control Options.
This section covers the following topics:
■
Other Shooting Newton Analyses
■
SN Error and Warning Messages

Other Shooting Newton Analyses

The following Shooting Newton Analyses are also supported by HSPICE
advanced analog analyses.
■
.SNFT is equivalent to the .FFT command in transient (.TRAN) analysis.
.SNFT uses Fourier transform to represent a time domain signal in the
frequency domain. For more information, see Shooting Newton with Fourier

62 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 3: Shooting Newton Analysis
Troubleshooting Convergence Issues with Shooting Newton Analysis

Transform (.SNFT).
■
.SNAC is used to perform a linear analysis of a driven (or non-autonomous)
circuit, where the linear coefficients are modulated by a periodic,
steady-state signal. The functionality is similar to the .HBAC command. For
more information, see Shooting Newton AC Analysis (.SNAC).
■
.SNXF is used to calculate transfer functions from an arbitrary number of
small signal sources to a designated output in a circuit under periodic
steady-state conditions. For more information, see Shooting Newton
Transfer Function Analysis (.SNXF).

SN Error and Warning Messages

Error messages are displayed with convergence recommendations in cases of
non-convergence within the maximum number of Shooting-Newton iterations.
Error messages are displayed for software errors such as segmentation
violations, and abort conditions such as:
■
unrecognized format such as unrecognized V/I source.
■
faulty input values such as wrong sign and out of range values.
■
unspecified values such as unspecified tone.
■
inconsistent values such as non-commensurable tones.
■
duplicate values such as same entry provided more than once, the last one
is always taken.

Troubleshooting Convergence Issues with Shooting

Newton Analysis
To troubleshoot convergence issues with shooting newton analysis, perform the
following steps:
■
Increase the accuracy using either the RUNLVL (default is 3) or the
SIM_ACCURACY (default is 10) option.
■
Increase the initial transient simulation time by increasing the value of
TRINIT in the .SN or the .SNOSC command.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 63

N-2017.12
Chapter 3: Shooting Newton Analysis
Troubleshooting Convergence Issues with Shooting Newton Analysis

■
Limit the maximum time-step size used in the simulation by setting the
option DELMAX to a value equal to the waveform period divided by 100.
■ Ensure that the correct frequency is selected. In the case of a driven circuit
containing dividers, use the final divider (slowest) output frequency.

64 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Part 2: Oscillator and PLL Analysis

The following chapters/topics are included in this Part:

■
Chapter 4, Oscillator Analysis
■ Chapter 5, Phase Noise Analysis
■ Chapter 6, Behavioral PLL Noise Analysis (.ACPHASENOISE)

HSPICE® User Guide: Advanced Analog Simulation and Analysis 65

N-2017.12
66 HSPICE® User Guide: Advanced Analog Simulation and Analysis
N-2017.12
 4
Oscillator Analysis
4

Describes how to use the HSPICE advanced analog functions to perform

oscillator and phase noise analysis on oscillator circuits.

Two main groups categorize oscillators:

■
Ring oscillators: These oscillators tend to have a low Q and operate based
on the delay of digital cells such as inverters. Ring oscillators have a strong
nonlinear behavior and the output signals are often square-wave-like. You
can analyze ring oscillators in either the frequency domain using harmonic
balance analysis or in the time domain using Shooting Newton analysis.
■
Harmonic oscillators: Common harmonic oscillators are LC and crystal
oscillators. These oscillators tend to have a high Q, making it difficult to find
the oscillation frequency. Their behavior tends to be only mildly nonlinear
and their output signals tend to be close to purely sinusoidal. harmonic
balance analysis is most effective for analyzing harmonic oscillators.
HSPICE advanced analog analyses includes special analysis algorithms for
finding the steady-state solution for oscillator circuits. No driving sources set
the frequencies of operation in oscillators. The fundamental oscillation
frequency is one of the unknowns that the simulator solves for. HSPICE
advanced analog analyses provide two approaches: harmonic balance
analysis or a Shooting Newton algorithm-based analysis.
This section covers the following topics:
■
Oscillator Analysis (.HBOSC and .SNOSC)
■
Shooting Newton Oscillator Analysis (.SNOSC)
■ Ring Oscillator Example
■
References

HSPICE® User Guide: Advanced Analog Simulation and Analysis 67

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Oscillator Analysis (.HBOSC and .SNOSC)

Because the frequencies of the driving sources do not determine the frequency
of oscillation, the simulator solves for a slightly different set of nonlinear
equations as in the following equation:

Equation 16
F  V  0  = I  V  0  + Q  V  0  + Y   0 V + I s
HSPICE harmonic balance oscillator analysis (.HBOSC) adds the fundamental
frequency of oscillation to the list of unknown circuit quantities. To
accommodate the extra unknown, HSPICE sets the phase (or equivalently, the
imaginary part of one unknown variable — generally a node voltage) to zero.
The phases of all circuit quantities are relative to the phase at this reference
node (referred to as the “PROBENODE”).
Additionally, the HBOSC analysis tries to avoid the “degenerate solution,”
where all non-DC quantities are zero. Although this is a valid solution of the
above equation (it is the correct solution, if the circuit does not oscillate), the
HBOSC analysis might find this solution incorrectly, if the algorithm starts from
a bad initial solution.
The HBOSC analysis follows a technique similar to that described by Ngoya et
al., which uses an internally-applied voltage probe to find the oscillation voltage
and frequency. The source resistance of this probe is a short circuit at the
oscillation frequency, and an open circuit otherwise. HSPICE advanced analog
analyses use a two-tier Newton approach to find a non-zero probe voltage
which results in a zero probe current.
HB analysis of the oscillator circuit uses the DC solution as a starting point.
This analysis requires, in addition to the DC solution, initial values for both the
oscillation frequency and the probe voltage. HBOSC analysis calculates the
small-signal admittance that the voltage probe sees over a range of
frequencies, in an attempt to find potential oscillation frequencies. Oscillation is
likely to occur where the real part of the probe current is negative and the
imaginary part is zero. You can use the FSPTS parameter to specify the
frequency search. You must also supply an initial guess for the large signal
probe voltage. A value of one-half the supply voltage is often a good starting
point.
For more information on control options, see the.HBOSC command in the
HSPICE Reference Manual: Commands and Control Options.

68 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

The following sections discuss these topics:

■
Tutorial Examples Using HBOSC Analysis
■
.HBOSC Output Syntax
■
Using the .MEASURE Command with .HBOSC
■
HB Simulation of Ring Oscillators
■
Troubleshooting Convergence Problems

Tutorial Examples Using HBOSC Analysis

The following tutorial examples illustrate:
■
Example 1 — Colpitts Oscillator
■
Example 2 — Using HBOSC for a CMOS GPS VCO

Example 1 — Colpitts Oscillator

This section demonstrates the HSPICE advanced analog oscillator analysis by
using a single transistor oscillator circuit. Oscillator analysis is an extension of
harmonic balance in which you solve for the base frequency. In oscillator
analysis, the user supplies a guess of the base frequency. It requires no voltage
or current source stimulus.
To activate oscillator analysis, include a .HBOSC command with:
■
The TONE parameter set to a guess of the oscillation frequency.
■
The PROBENODE parameter set to identify an oscillating node or a pair of
nodes. Always specify a pair of nodes; if only one node oscillates, specify
ground as the second node. To speed up the simulation, you also supply a
guess of the magnitude of the oscillating voltage across these nodes.
■ The FSPTS parameter set to a frequency range and the number of search
points. When you set FSPTS, HSPICE advanced analog analyses precede
the HBOSC analysis, with a frequency search in the specified range to
obtain an optimal initial guess for the oscillation frequency. This can
accelerate the HB oscillator convergence.
In conjunction with oscillator analysis, HSPICE advanced analog analyses can
perform phase noise analysis. Phase noise analysis measures the effect of
transistor noise on the oscillator frequency. You activate phase noise analysis
by using the .PHASENOISE command; this command sets a set of frequency

HSPICE® User Guide: Advanced Analog Simulation and Analysis 69

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

points for phase noise analysis. The .PRINT and .PROBE commands output
the phase noise values.
The following netlist, osc.sp, simulates an oscillator and performs phase
noise analysis. This example file is available in the HSPICE distribution in the
directory $installdir/demo/hspice/rf_examples/.

Figure 10 Colpitts Oscillator

Use the .HBOSC command with the PROBENODE and FSPTS parameters set.
PROBENODE=emitter,0,4.27

Identifies the emitter node as an oscillating node and provides a guess value of
4.27 volts for the oscillation amplitude at the emitter node.
FSPTS=40,9e6,1.1e7

Causes an initial frequency search using 40 equally-spaced points between 9

and 11 MHz.
In the .PHASENOISE, .PRINT, and .PROBE commands:

70 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

.PHASENOISE V(emitter) dec 10 10k 1meg

Runs phase noise analysis at the specified offset frequencies, measured from
the oscillation carrier frequency. The frequency points specified here are on a
logarithmic scale, 10 points per decade, 10 kHz to 1 MHz.
■
.PROBE PHASENOISE PHNOISE and the similar .PRINT command
instruct HSPICE advanced analog analyses to output phase noise results to
the osc.pn0 and osc.printpn0 files.
**
** Uses emitter resistor limiting to keep output sinusoidal.
** Output can be taken at the emitter (eml node).
**
*---------------------------------------------------------
* Options for Oscillator Harmonic Balance Analysis...
*
.OPTIONS post sim_accuracy=100 hbsolver=0
*---------------------------------------------------------
* Bias NPN transistor for 5V Vce, 10mA Ic
* Emitter follower Colpitts design
Vcc collector 0 9V
Q1 collector base emitter emitter RF_WB_NPN
Re1 emitter eml 100
RLoad eml 0 300
Rb1 collector base 4300
Rb2 base 0 5600
*
*---------------------------------------------------------
* Capacitive feedback network
Ce 0 eml 100pF
Cfb base eml 100pF
Cbb base bb 470pF
Lb bb 0 6uH
*---------------------------------------------------------
* Simulation control for automated oscillator analysis
*
.HBOSC tones=1.0e7 nharms=15
+PROBENODE=emitter,0,4.27
+FSPTS=40,9.e6,1.1e7
*
.PHASENOISE V(emitter) DEC 10 10K 1MEG
+METHOD=0 CARRIERINDEX=1
*
.print hbosc vm(eml) vp(eml) vr(emitter) vi(emitter)
.print hbosc vm(emitter) vp(emitter) P(Rload)
.print phasenoise phnoise
.probe phasenoise phnoise

HSPICE® User Guide: Advanced Analog Simulation and Analysis 71

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

.probe hbosc v(emitter) v(eml)

.include bjt.inc
.END

After you run this netlist, examine the osc.printhb0 file.

■
At the top is the oscillator frequency (about 10.14 MHz) and the .PRINT
HBOSC output.
■
The first two lines show that the eml node oscillates around 3 V with an
amplitude of about 2.85 V.
■
The emitter node oscillates around 4V with an amplitude of about 4.27 V.
Also examine the osc.printpn0 file which contains the phase noise results
in text form.
You can view the osc.hb0 and osc.pn0 files in Custom WaveView.
1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the osc.hb0 file from the
Open: Waveform Files dialog box.
3. Select the v(emitter) signal in the signal browser. Double-click on the
signal name or drag and drop the signal in the waveform.

72 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

4. In the waveform, right-click in the name area of the panel containing the
signal v(emitter), left-click on the waveform label for v(emitter) in the
waveform. From the Panel menu, choose Signal 'v(emitter)' > To Time-
Domain.
5. To accept the defaults for range and interval, click OK in the Convert to Time
domain window.
6. In a new waveform, you should now see a time domain waveform named
IFT.0|v(emitter).

HSPICE® User Guide: Advanced Analog Simulation and Analysis 73

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

To run a transient simulation for comparison:

1. Use the .TRAN 1n 10u command.
2. Add ic=10n to the Lb inductor.
The resulting waveforms should be the same as those from the HB oscillator
analysis.
Time domain waveforms can also be output from HB analysis. In the netlist,
include the following .PROBE statement:
.probe hbtr v(emitter)

This will output a osc.hr0 file that will contain the oscillator output waveform in
the time domain.

74 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Example 2 — Using HBOSC for a CMOS GPS VCO

This second oscillator analysis example involves two negative resistance
oscillators coupled at 90 degrees. The MOS capacitors double as varactors.
This VCO topology is common for GPS applications and produces quadrature
LO outputs near 1550 MHz. The purpose of this example is to generate the
VCO tuning curve (output level and frequency, as a function of the tuning
voltage), as well as its phase noise characteristics as a function of tuning
voltage.
You activate the oscillator analysis by using the .HBOSC command:
■
The TONE parameter sets an oscillation frequency (near 1550 MHz).
■
The NHARMS parameter sets the harmonic content to the 11th order.
■ The PROBENODE parameters identify the drain pins across the first oscillator
section as the pair of oscillating nodes. This is a differential oscillator and
the approximate value for this differential amplitude is 6.1 V.
■ The FSPTS parameters set the search frequency range between 1500 and
1600 MHz.
■
The SWEEP parameters set a tuning voltage sweep from 2.0 to 3.2 V.
The following example uses the demonstration netlist gpsvco.sp, which is
available in the directory $installdir/demo/hspice/rf_examples/.
This netlist simulates the oscillator schematic Figure 11 and performs phase
noise analysis.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 75

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Figure 11 VCO Schematic

76 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

**
** NMOS IC Quadrature VCO circuit for GPS local oscillator
**
** Twin differential negative resistance VCOs using NMOS
** transistors for varactors, coupled to produce quadrature
** resonances.
** Design based on 0.35um CMOS process.
**
** References:
** >P. Vancorenland and M.S.J. Steyaert, "A 1.57-GHz fully
** integrated very low-phase-noise quadrature VCO,"
** IEEE Trans. Solid-State Circuits, May 2002, pp.653-656.
** >J. van der Tang, P. van de Ven, D. Kasperkovitz, and A.
Roermund,
** "Analysis and design of an optimally coupled 5-GHz quadrature
** LC oscillator," IEEE Trans. Solid-State Circuits, May 2002,
** pp.657-661.
** >F. Behbahani, H. Firouzkouhi, R. Chokkalingam, S. Delshadpour,
** A. Kheirkhani, M. Nariman, M. Conta, and S. Bhatia,
** "A fully integrated low-IF CMOS GPS radio with on-chip analog
** image rejection," IEEE Trans. Solid-State Circuits, Dec. 2002,
** pp. 1721-1727.
** Setup for Harmonic Balance Analysis
** Oscillation Frequency: ~ 1575 MHz (GPS L1 frequency)
** Amplitude: ~5 Volts peak-to-peak (zero to 5V)
** Vdd: 2.5 V
**
** HSPICE Simulation Options:
*.option delmax=1n ACCURATE LIST NODE
**
** HSPICE advanced analog simulation options :
.option sim_accuracy=10
**
*.option savehb=’a.hbs’ loadhb=’a.hbs’
.option POST
.param Vtune=2.0 $ Failures: vtune=1
.param Cval=0.2p
*---------------------------------
Vtune vc gnd DC Vtune
Vdd vdd gnd 2.5
*---------------------------------
* First oscillator section
** Low-Q resonator with Vdd at center tap of inductors
R1a IP ri 100k $ These R’s set the Q
R1b ri IN 100k
L1 IP vdd 16.5nH
L2 vdd IN 16.5nH
Cc1 IP gnd Cval $ I to Q

HSPICE® User Guide: Advanced Analog Simulation and Analysis 77

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Cc2 IN gnd Cval $ -I to Q

** Differential fets
M1 IP IN cs gnd NMOS l=0.35u w=15u
M2 IN IP cs gnd NMOS l=0.35u w=15u
** Bias fet - bias at Vdd -- too high?
Mb cs vdd gnd gnd NMOS l=0.35u w=15u
** fets used as varactors
Mt1 vc IP vc gnd NMOS l=0.35u w=2u M=50
Mt2 vc IN vc gnd NMOS l=0.35u w=2u M=50
*---------------------------------
** Second oscillator section
** Low-Q resonator with Vdd at center tap of inductors
R1a_b QP ri_b 100k $ These R’s set the Q
R1b_b ri_b QN 100k
L1_b QP vdd 16.5nH
L2_b vdd QN 16.5nH
Cc1_b QP gnd Cval $ -Q to -I
Cc2_b QN gnd Cval $ -Q to I
** Differential fets
M1_b QP QN cs_b gnd NMOS l=0.35u w=15u
M2_b QN QP cs_b gnd NMOS l=0.35u w=15u
** Bias fet - bias at Vdd -- too high? 2nd in parallel
Mb_b cs_b vdd gnd gnd NMOS l=0.35u w=15u
** fets used as varactors
Mt1_b vc QP vc gnd NMOS l=0.35u w=2u M=50
Mt2_b vc QN vc gnd NMOS l=0.35u w=2u M=50
*
*-------------------------------
* Differentiators Coupling transistors for quadrature
*
.param Cdiff=0.14p difMsize=50u
vidiff dbias gnd 1.25
viqdiff vdcdif gnd 1.75
Midiff1 dQP dbias gnd gnd NMOS l=0.35u w=difMsize
Midiff2 dQN dbias gnd gnd NMOS l=0.35u w=difMsize
Midiff3 dIN dbias gnd gnd NMOS l=0.35u w=difMsize
Midiff4 dIP dbias gnd gnd NMOS l=0.35u w=difMsize
Cdiff1 dQP QP Cdiff
Cdiff2 dQN QN Cdiff
Cdiff3 dIN IN Cdiff
Cdiff4 dIP IP Cdiff
Mc_QP1 IP vdcdif dQP gnd NMOS l=0.35u w=difMsize
Mc_QN2 IN vdcdif dQN gnd NMOS l=0.35u w=difMsize
Mc_QN3 QP vdcdif dIN gnd NMOS l=0.35u w=difMsize
Mc_QP4 QN vdcdif dIP gnd NMOS l=0.35u w=difMsize
*-------------------------------
* Transient Analysis Test Bench
* stimulate oscillation with 2mA pulse

78 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

*iosc IP IN PULSE ( 0 2m .01n .01n .01n 10n 1u )

*.probe tran v(IP) v(IN)
*.print tran v(IP) v(IN)
*.TRAN .01n 10n
*------------------------------
* Harmonic Balance Test Bench
*
.sweepblock vtune_sweep
+ 0 5 0.2
+ 2 3 0.1
.HBOSC tones=1550e6 nharms=12
+ PROBENODE=IP,QN,4
+ sweep Vtune sweepblock=vtune_sweep
**
.phasenoise v(IP,IN)dec 10 100 1e7
.print phasenoise phnoise
.probe phasenoise phnoise
.print hb v(IP,IN) v(IP,IN)[1] v(QP,QN) v(QP,QN)[1]
.probe hb v(IP,IN) v(IP,IN)[1] v(QP,QN) v(QP,QN)[1]
.probe hbtr v(IP,IN) v(IP,IN)
.probe hb hertz[1]
*
.END

Viewing Results Using Custom WaveView

You can view the VCO frequency and time domain waveforms, tuning curves,
and phase noise response using Custom WaveView.
To view the VCO frequency domain waveform:
1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the gpsvco.hb0 file.
3. Select the signals v(qp,qn) and v(ip,in) in the signal browser. Drag
and drop the signals in the waveform.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 79

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Figure 12 VCO Spectra Output

To view the time domain waveforms:

1. Use File > Import Waveform File and select the gpsvco.hr0 file.
2. Select the signals v(qp,qn) and v(ip,in) in the signal browser. Drag
and drop the signals in the waveform.

80 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Figure 13 VCO Waveform Output

To view the tuning curves of the VCO:

1. Use Waveview > New to open a new waveform.
2. Select the signal par(hertz[1]):(1) in the signal browser. Drag and
drop the signal in the waveform. This will show the change in frequency as
the tuning voltage is changed from 0 to 5 volts.
3. Select the signal v(qp,qn):(1) in the signal browser. Drag and drop the
signal into the same waveform as the previous signal. This will show the
change in magnitude of the signal as the tuning voltage is changed.
4. Change to the signal mode by clicking on the signal mode icon in the
waveform icon bar.
5. Select the v(qp,qn):(1) signal in the waveform. The signal will now be
highlighted.
6. Right-click on the highlighted signal. In the menu, scroll down to
Highlighted Signals. A menu will appear. Scroll down to Right Axis and
click on it. The right axis will now be the magnitude of the signal
v(qp,qn):(1).
7. To show the data points, click on the toggle data points icon in the icon bar.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 81

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

To view the phase noise analysis results:

1. Use File > Import Waveform File and select the gpsvco.pn0 file.
2. Use Waveview > New to open a new waveform.
3. Select the signal nlp_l(f) in the signal browser. Drag and drop the signal
in the waveform.
4. The X-axis scale will be linear and it is preferred to use the log scale. To
change the X-axis scale to the log scale, right-click the X-axis. A menu will
appear. Select log in this menu.

82 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Figure 14 VCO Phase Noise Response

.HBOSC Output Syntax

The output syntax for .HBOSC analysis is identical to that for HB analysis (see
Chapter 2, Harmonic Balance Analysis). To output the final frequency of
oscillation, use the HERTZ keyword. For example, HERTZ[1] identifies the
fundamental frequency of oscillation.

Note: For PROBENODE = n1 n2 vp, where vp is a voltage, the units

must be in volts.

See also Outputting Phase Noise Source as ASCII Data Files Using *.printpn0.

Using the .MEASURE Command with .HBOSC

Since .HBOSC requires an .HB analysis, the measure statements for this
analysis are the same as for .HB analysis. For example,
.MEASURE HB result FIND out_var AT=val

HSPICE® User Guide: Advanced Analog Simulation and Analysis 83

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

HB Simulation of Ring Oscillators

Ring oscillators require a slightly different simulation approach in HB analysis.
Since their oscillation is due to the inherent delay in the inverters of the ring,
they are best modeled in the time domain and not in the frequency domain.
In addition, ring oscillator waveforms frequently approach square waves, which
require a large number of harmonics in the frequency domain. An accurate
initial guess is important for an accurate HB simulation.
The HSPICE advanced analog HBOSC analysis typically starts from the DC
solution and looks for potential resonances in the linear portion of the circuit to
determine the initial guess for the oscillation frequency. However, these
resonances generally do not exist in ring oscillators which do not contain linear
resonant elements.
HB analysis provides a second method of obtaining a good initial guess for the
oscillation frequency, specifically intended for ring oscillators. Instead of
starting from the results of a DC analysis, this method starts from the result of a
transient analysis. This method, Transient Initialization, also provides a good
initial guess for all the voltages and currents in the circuit.
The recommended setup for ring oscillators is therefore:
■
Set up .HBOSC without FSPTS.
■
Choose one of the nodes in the ring as the PROBENODE.
■
Recommendation: Since ring oscillators tend to have square-wave-like
output signals which have significant high frequency content, use a
relatively large value, perhaps 50, for nharms. Ring oscillators with more
stages tend to need more harmonics.
■
Set HBTRANINIT to a value that represents ~5-10 oscillator periods. Ensure
that you include an .ic command or other transient analysis setup to start
the oscillator in transient simulation. Longer HBTRANINIT times may result
in faster HBOSC convergence, at the expense of additional CPU time spent
on HBTRANINIT.

Troubleshooting Convergence Problems

This section lists the most common causes of convergence problems and how
to recognize them and resolve them.

84 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

The HSPICE advanced analog harmonic balance oscillator analysis consists of

a two-tier iterative analysis of inner loop and outer loop iterations. In the outer
loop iteration, HBOSC iterates to reduce any reported “probe errors” for each
outer loop iteration. Each outer loop iteration involves a non-autonomous HB
circuit solution; this non-autonomous solve is the inner loop iteration.
If HBOSC has inner loop convergence problems, the simulation may hang on
the first outer loop iteration or you may see warning messages such as:
Warning: HB_WARN.3: Final HB residual value > HB_TOL.
Rank of HB Jacobian = 155
Warning: HB_WARN.3: HB convergence failure in non-autonomous HB.

The simulation lists the probe voltage and probe frequency for each outer loop
iteration. If an outer loop convergence problem occurs, you may see the
following:
■
Decreasing probe voltage values.
■
Wildly fluctuating values of probe frequency.
Osc probe : voltage = 0.218234 frequency =
6.240794122744832e+09
■
A warning message which indicates that the oscillator simulation has
reached a non-oscillating DC solution.
Warning: HB_ERR.18: HB oscillator analysis has reached the
NULL solution.

The following sections discuss these topics:

■
General Convergence Issues
■
Outer Loop Convergence
■
Inner Loop Convergence

General Convergence Issues

The following sections discuss these topics:
■
Probe Node Location
■
Incorrect Source Values
■
GMRES Convergence
■
Accuracy of Initial Guess

HSPICE® User Guide: Advanced Analog Simulation and Analysis 85

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Probe Node Location

Since convergence is sensitive to the probe node location, you can often track
convergence problems to this setting.
A common scenario is that the oscillator's output signal passes through one or
more buffers and a designer may think of the buffer output as the oscillator
output. A frequent mistake places the probe node at the output of the buffer, but
this causes HB convergence problems. In this case, move the probe node to
the oscillator part of the circuit. Often, it is necessary to select an internal node
of a subcircuit to achieve this.
The most typical symptom of this problem is inner loop convergence failure. For
ring oscillators, always choose a node that is part of the ring, that is, connecting
two stages of the ring; for harmonic oscillators, choose a node close to the
oscillator.

Incorrect Source Values

If the original netlist simulates the oscillator in transient analysis, some voltage
or current sources may have transient descriptions (for example, PWL) to start
the oscillator. For example, you can ramp a voltage supply to simulate a power-
up to start the oscillator:
Vvdd vdd 0 PWL (0 0 1n 3)
In this case, the user would like HBOSC to use 3 as the voltage source value,
but HBOSC uses 0 because harmonic balance uses the explicit DC value of
the source. HSPICE advanced analog analyses tries to interpret your sources
intelligently but, in some cases, it may not be able to determine what you
intended.
For the above example, there are a few ways to ensure that HSPICE advanced
analog analyses correctly interprets the source.
■
Remove the explicit DC value. If you only provide a transient description, HB
uses the time=infinity value of the source.
■
Add TRANFORHB=1
Vvdd vdd 0 PWL (0 0 1n 3) TRANFORHB=1

The TRANFORHB=1 keyword causes HB to use the transient analysis

description in HB and HBOSC.
■
Add an explicit HB value

86 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

Vvdd vdd 0 PWL (0 0 1n 3) HB 3 0 0

This causes HB to treat the source as a 3V DC source since the HB value

specifies the 0th harmonic). If you provide an HB value, HSPICE advanced
analog analyses ignore the PWL description and uses “HB 3 0 0”
(amplitude=3, phase=0, harmonic=0) instead. The PWL description still
applies for HBTRANINIT.
Incorrect source values usually result in the following:
■ High residual value after HBTRANINT. Usually, HBTRANINIT should
produce a good starting point for HB or HBOSC. Typical residuals after
HBTRANINIT are 1e-4 or 1e-5. If the initial residual, printed immediately
after HBTRANINIT completes, is high, there may be a source problem. In
the VDD ramping example above, you might see a residual value of 3. Outer
loop may converge to a DC solution because incorrect source values result
in a non-oscillatory circuit:
Warning: HB_ER.18: HB oscillator analysis has reached the NULL
solution.
■ In some cases, inner loop non-convergence may occur.

GMRES Convergence
HSPICE advanced analog analyses use a GMRES iterative solver to solve the
linear systems that arise on each inner loop Newton-Raphson step. If GMRES
does not solve the linear systems accurately enough, then the inner loop may
not converge.
The GMRES solver is controlled by two options:
■ HBKRYLOVTOL: relative tolerance for GMRES solver. Default is 0.01, or 1%.
For some circuits, setting this option helps inner loop convergence:
.option HBKRYLOVTOL=1e-3
■
HBKRYLOVDIM: dimension of Krylov subspace to use in GMRES iteration.
Also controls maximum number of GMRES iterations. The HSPICE
advanced analog analyses .lis file lists the number of GMRES iterations
taken for each Newton-Raphson step. If that number is equal to
HBKRYLOVDIM, you may improve convergence by increasing
HBKRYLOVDIM. For example:
.option HBKRYLOVDIM=80

HSPICE® User Guide: Advanced Analog Simulation and Analysis 87

N-2017.12
Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

The symptom for GMRES convergence difficulty is always inner loop

convergence failure, or slow inner loop convergence. If this problem occurs, the
inner loop convergence is often good until the residual reaches a fairly low
value like 1e-8 or 1e-7, and then stagnates.

Accuracy of Initial Guess

Both inner loop and outer loop convergence improves significantly if the
starting point or initial guess of the iterative method is good.

Outer Loop Convergence

For outer loop convergence, the initial guess consists simply of the oscillation
frequency and the first harmonic amplitude at the probe node location. If inner
loop convergence is successful but outer loop convergence is not, then you
may need to provide a better frequency or amplitude guess.
If you use HBTRANINIT, then you can improve the accuracy of the initial guess
by use of one of the following methods:
■
Increase the HBTRANINIT time, simply by increasing the value of the
HBTRANINIT option.
■
Increase the HBTRANINIT accuracy. You can increase the transient
analysis accuracy by setting .option DELMAX or .option
SIM_ACCURACY. For example, you may set:
.option SIM_ACCURACY=10 HBTOL=1e-8

SIM_ACCURACY simultaneously tightens transient and HB accuracy

tolerances. If you want HB accuracy to remain unaffected, you may also
want to set HBTOL as in the example above.
■
Increase accuracy of the time domain to frequency domain conversion of
HBTRANINIT results, by increasing HBTRANPTS or equivalently,
decreasing HBTRANSTEP. For example:
.option HBTRANSTEP=1p

If you are using FSPTS, you can increase the number of points. Sometimes, it
is best to supply a guess manually by removing FSPTS and adjusting the
TONES value.
If you do not use HBTRANINIT, you may be able to improve convergence by
manually adjusting the PROBENODE amplitude guess.

88 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Oscillator Analysis (.HBOSC and .SNOSC)

To evaluate the effectiveness of your option settings, look at the “probe error”
reported after the first outer loop iteration:
Iteration 1
Osc probe : voltage = 0.2 frequency = 5.980000000000000e+09
hb residual = 7.628260e-10
Rank of HB Jacobian = 9102
Probe error = 0.000154462
dv = -0.0411324 df =-2.30464430e+08

A smaller probe error value indicates a better initial guess.

Inner Loop Convergence

If inner loop convergence is a problem, it may be because the initial voltage
waveform values are not close to the solution. The only way to improve the
voltage values is by using HBTRANINIT. While this does not work well for
harmonic oscillators, it does work well for ring oscillators. You can improve the
accuracy of HBTRANINIT as described in the outer loop convergence section
above.
If the initial residual is large after HBTRANINIT, you may want to check to make
sure that the voltage and current sources are consistent between HB and
transient analysis.
The following section discuss these topics:
■
Insufficient Number of Harmonics
■
Presence of Frequency Divider

Insufficient Number of Harmonics

If the number of harmonics specified is too small to represent the signals
present in the circuit, you may see convergence problems in either the inner or
the outer loop, or the solution may converge to an unreasonable frequency
value.
It is difficult to know when the number of harmonics is insufficient, but if you
suspect an insufficiency, it is a simple experiment to increase the value of
NHARMS. If you achieved convergence and the number of harmonics is large
enough, then the magnitude of the spectral data for all signals should
significantly decay with increasing frequency. If the spectral data for node
voltages has not decayed at the highest harmonics included in the simulation,
increase the value of NHARMS.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 89

N-2017.12
Chapter 4: Oscillator Analysis
Shooting Newton Oscillator Analysis (.SNOSC)

Presence of Frequency Divider

If a frequency divider is present and not accounted for by the SUBHARMS
setting, convergence is not possible because the Harmonic Balance spectrum
does not include the necessary low frequency components. As a result, inner
loop convergence fails. When debugging HBOSC convergence problems, it is
necessary to rule out the possibility of the presence of frequency dividers early
in the process.
If a frequency divider is present, you can simulate the circuit if you set
SUBHARMS to the largest frequency division present in the circuit. If a
frequency divider is present, it is almost always necessary to use the
HBTRANINIT option to achieve convergence.

Shooting Newton Oscillator Analysis (.SNOSC)

The analysis described in Chapter 3, Shooting Newton Analysis also provides a
very effective means for finding the steady-state for oscillator circuits.
Ring oscillators are best suited for time domain analysis by using Shooting
Newton because they tend to:
■
have a low Q
■
operate based on digital delays
■
have a strongly nonlinear behavior
■
output signals that are piece-wise-linear or square-wave-like
HBOSC is superior for sinusoidal waveforms. As with the Harmonic Balance
approach, the goal is to solve for the additional unknown oscillation frequency.
Shooting Newton accomplishes this by considering the period of the waveform
as an additional unknown, and solving the boundary conditions at the
waveform endpoints that coincide with the steady-state operation. As with
regular Shooting Newton analysis, you can specify the input in terms of time or
frequency values.
For more information on the control options, see the .SNOSC command in the
HSPICE Reference Manual: Commands and Control Options.
.SNOSC Output Syntax
The output syntax for .SNOSC analysis is identical to that for SN analysis (see
Chapter 3, Shooting Newton Analysis). To output the final frequency of

90 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Ring Oscillator Example

oscillation, use the HERTZ keyword. For example, HERTZ[1] identifies the
fundamental frequency of oscillation.
See also Using Noise Analysis Results as Input Noise Sources.

Ring Oscillator Example

The Shooting Newton algorithm provides a fast and effective analysis for ring
oscillators. The ringoscSN.sp input file for this example is located at
$installdir/demo/hspice/rf_examples/.

Figure 15 Ring Oscillator

.title ringosc

.subckt inv in out vdd

mn1 out in 0 0 nmos l=0.25u w=2u
mp1 out in vdd vdd pmos l=0.25u w=6u
.ends

vdd vdd 0 3
x1 1 2 vdd inv
x2 2 3 vdd inv
x3 3 4 vdd inv
x4 4 5 vdd inv
x5 5 6 vdd inv
x6 6 7 vdd inv
x7 7 1 vdd inv
c1 1 0 0.022p
.ic v(1)=3
.options post
.options snaccuracy=50

.snosc tones=335meg nharms=10 oscnode=1 trinit=10n

HSPICE® User Guide: Advanced Analog Simulation and Analysis 91

N-2017.12
Chapter 4: Oscillator Analysis
Ring Oscillator Example

.phasenoise v(7) dec 10 100 10meg

.probe sn v(7)
.probe snfd v(7)
.print phasenoise phnoise v(7)
.probe phasenoise phnoise v(7)

.end

This analysis finds the oscillation frequency of the ring oscillator. Since the
circuit is an oscillator, no input source is required. The oscillator is started by
setting an initial condition at the input of the ring (node 1). In the .SNOSC
command, the node that the analysis will probe for oscillation conditions is
specified, as well as the approximate frequency of oscillation. The number of
harmonics to include in the analysis is specified as well.
The phase noise characteristics of the oscillator are analyzed by using the
.PHASENOISE command. The .PHASENOISE command requires that an
output node, a pair of nodes, or a two-port element and a frequency sweep be
specified. The frequency sweep is used to calculate the phase noise analysis at
the specified offset frequencies, measured from the oscillation carrier
frequency. For this example of phase noise analysis, the default Nonlinear
Perturbation (NLP) method is used.
The signal v(7) will be probed in both the frequency and time domain. The
measure statement is used to measure the fundamental frequency of the
oscillator.
Simulation Status Output
During the simulation, the simulation status is displayed on the screen. In
addition to the screen display, a more detailed status, CPU time, and memory
usage information are also written to the ringoscSN.lis file.
Viewing Results in Custom WaveView
You can view the time-domain ringoscSN.sn0 file, the frequency domain
ringoscSN.snf0 file, and the phase noise ringoscSN.snpn0 file in
Custom WaveView.
1. Enter wv at the prompt to start Custom WaveView.
2. Use File > Import Waveform File to open the ringoscSN.sn0 file.
3. Select the signal v(7) from the signal browser. Drag and drop the signal v(7)
to the right side of the waveform so that panels are opened in row / column
format. The time domain trace shown at the right side of Figure 16 on
page 93.

92 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
Ring Oscillator Example

4. Use File > Import Waveform File to open the ringoscSN.snf0 file.
5. Select the signal v(7) from the signal browser. Drag and drop the signal v(7)
to the right side of the waveform so that panels are opened in row / column
format. The frequency domain spectrum is shown at the left side of
Figure 16 on page 93.

Figure 16 Ring Oscillator Output

6. Use File > Import Waveform File to open the ringoscSN.snpn0 file.
7. Use Waveview > New to open a new waveform.
8. Select the signal nlp_l(f) from the signal browser. Drag and drop the signal
nlp_l(f) signal in the waveform. Figure 17 on page 94 shows the resulting
phase noise results for the oscillator.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 93

N-2017.12
Chapter 4: Oscillator Analysis
Ring Oscillator Example

Figure 17 Ring Oscillator Phase Noise

94 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 4: Oscillator Analysis
References

References
[1] E. Ngoya, A. Suarez, R. Sommet, R. Quere, “Steady State Analysis of Free
or Forced Oscillators by Harmonic Balance and Stability Investigation of
Periodic and Quasi-Periodic Regimes,” International Journal of Microwave
and Millimeter-Wave Computer-Aided Engineering, Volume 5, Number 3,
pages 210-223 (1995)
[2] C.R. Chang, M.B. Steer, S. Martin, E. Reese, “Computer-Aided Analysis of
Free-Running Microwave Oscillators,” IEEE Trans. on Microwave Theory
and Techniques, Volume 39, No. 10, pages 1735-1745, October 1991.
[3] G.D. Vendelin, Design of Amplifiers and Oscillators by the S-Parameter
Method, John Wiley & Sons, 1982
[4] A. Demir, A. Mehrotra, J. Roychowdhury, “Phase Noise in Oscillators: A
Unifying Theory and Numerical Methods for Characterization” in Proc. IEEE
DAC, pages 26-31, June 1998.
[5] A. Demir, A. Mehrotra, and J. Roychowdhury, “Phase Noise in Oscillators: A
Unifying Theory and Numerical Methods for Characterization,” IEEE Trans.
Circuits System I, Volume 47, pages 655–674, May 2000.
[6] A. van der Ziel, Noise in Solid State Devices and Circuits, John Wiley &
Sons, 1986.
[7] A. Hajimiri, S. Limotyrakis, and T.H. Lee, “Jitter and phase noise in ring
oscillators,” IEEE J. Solid-State Circuits, vol. 34, no. 6, pp. 790-804, June
1999.
[8] Jitter Analysis Techniques for High Data Rates, Application Note 1432,
Agilent Technologies, Feb. 2003.
[9] Characterization of Clocks and Oscillators, NIST Technical Note 1337,
National Institute of Standards and Technology.
[10] G.V. Klimovitch, “Near-carrier oscillator spectrum due to flicker and white
noise,” Proc. ISCAS 2000 (Geneva), May 2000.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 95

N-2017.12
Chapter 4: Oscillator Analysis
References

96 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 5
5 Phase Noise Analysis

Describes phase noise analysis.

Phase Noise analysis requires first running either the harmonic balance
(HBOSC) or Shooting Newton (SNOSC) analysis, and then PHASENOISE
analysis. The PHASENOISE analysis itself is identical whether you run
SNOSC or HBOSC.
For more information on control options, see the .PHASENOISE command in
the HSPICE Reference Manual: Commands and Control Options.
This chapter discusses the following topics:
■
Introduction to Phase Noise Analysis (.PHASENOISE)
■
Accumulated Jitter Measurement for Closed Loop PLL Analysis
■ Clock Source with Random Jitter

Introduction to Phase Noise Analysis (.PHASENOISE)

The following figure shows a simple free-running oscillator, which includes a
port with injected current.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 97

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

in
v(t)

Figure 18 Oscillator with Injected Current

The following equation presents how an ideal oscillator would be insensitive to

perturbations with a fixed amplitude and frequency:

Equation 17
v  t  = A cos   0 t +  0 
A noisy oscillator has amplitude and phase fluctuations we can write as:

Equation 18
v  t  = A  t  cos   0 t +   t  
In the preceding equation:
■
A(t) is the time varying amplitude for the noisy oscillator.
■
  t  is the time varying phase for the noisy oscillator.
■
 0 is the frequency of oscillation.
In most applications, the phase noise is of particular interest, because it
represents frequency fluctuations about the fundamental, which you cannot
remove. These fluctuations are random processes. Typical expressions are in
terms of their power spectral density. For most oscillators, the phase noise is a
low-frequency modulation that creates sidebands in the oscillator’s spectrum,
about  0 .
For example, the following equation represents a simple sinusoidal variation in
the phase:

Equation 19
v  t  = A cos   0 t +  p sin  m t 

98 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

■
 p is the peak phase deviation, specified as  p =    m .
■
 is the peak angular frequency deviation.
For  p « 1 , the following equation approximates the output:

Equation 20
 p 
v  t  = A  cos   0 t  – -----  cos   0 +  p  t – cos   0 +  m  t  
 2 
That is, when the peak phase deviation is small, the result is frequency
components on each side of the fundamental with amplitude  p  2 .

The Single-Sideband Phase Noise L  f m  is the ratio of noise power to carrier

power in a 1Hz bandwidth, at offset  m = 2f m , which in this case can be
written as:

Equation 21
2 2
V sb 2 p rms
L  f m  =  -------- = ----------- = ----------------
A 4 2
This model for oscillator noise shows that sidebands about the fundamental,
due to noise, directly relate to the spectrum of the phase fluctuations   t  . The
more general definition of phase noise relates it to the spectral density of phase
fluctuations, that is:

Equation 22
 p2
S    m  = ----- = 2L  f 
2 m
HSPICE advanced analog analyses use several sophisticated analysis
techniques for computing the power spectrum of the phase variations to yield
the phase noise response. This information describes the spectrum of the
oscillator about the fundamental frequency, and its random jitter characteristics.
Any .PHASENOISE analysis results in the calculation of a curve fit for a
power-law model according to:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 99

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Equation 23
 a3 a2 a1 
L  f  = 10  log  -----3- + -----2- + ------ + a0 
f f f 

The .lis file reports the coefficients a3 , a2 , a1 , and a0 .

The .lis file includes a table which models the phase noise, including the
behavioral model fit and its fit error.
|-------------------------------------------------------------|
| L(f) = 10*log10( a3/f^(2+ef) + a2/f^2 + a1/f^(ef) + a0 ) dBc/Hz|
| a3 = 0.000000e+00 |
| a2 = 3.165111e-02 |
| a1 = 0.000000e+00 |
| a0 = 0.000000e+00 |
| ef = 1.000000e+00 |
| Average fit error = 1.6185e+00 dB |
| Maximum fit error = 8.4862e+00 dB @ 1.0000e+07 Hz |

The following sections discuss these topics:

■
Identifying Phase Noise Spurious Signals
■
Phase Noise Algorithms
■
PHASENOISE Output Syntax
■
Measuring Phase Noise with .MEASURE PHASENOISE
■
Optimizing with PHASENOISE Measurement
■
Amplitude Modulation/Phase Modulation Separation
■
Interpreting Phase Noise Analysis Results

Identifying Phase Noise Spurious Signals

Realistic phase noise responses include spurs. Spurs are contributions to the
phase noise that result from deterministic signals present within the circuit. In
most cases, the spurs are very small signals and do not interfere with the
steady-state operation of the oscillator, but do add energy to the output
spectrum of the oscillator. You may need to include the energy that the spurs
add in jitter measurements. The phase noise spurs feature adds an additional
analysis option that can predict the spurious contributions to the jitter.

100 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

To activate the new phase noise spur analysis, use the SPURIOUS keyword in
the .PHASENOISE command. An additional .HBAC analysis predicts the
spurious contributions to the phase noise.
The use of a voltage or current source can to add spurious signals to an
oscillator circuit. The keyword SPUR identifies the spurious signal.
Syntax
Exxxx n1 n2 … [SPUR mag phase freq] … $ voltage spur
Gxxxx n1 n2 … [SPUR mag phase freq] … $ current spur
where:
■
mag is the amplitude in volts or amps
■
phase is the phase in degrees
■
freq is the frequency in Hz
The source is equivalent to a steady-state sinusoidal source at the specified
amplitude, phase, and frequency values and only the spurious analysis uses it.
All other analyses ignore it. The SPUR keyword is combinable into a source
that other analyses use.
Recommendation: Add SPUR sources as separate sources.

Phase Noise Algorithms

HSPICE provides three approaches for oscillator phase noise analysis:
Nonlinear Perturbation (NLP), Periodic AC (PAC), and Broadband Phase Noise
(BPN) validation. Setting the METHOD parameter to 0, 1, or 2, selects the first,
second, and the third approach respectively. The three methods are provided
for flexible control of the simulation speed and accuracy. The NLP method is
the fastest, while the PAC method has superior accuracy over a wide range of
offset frequencies. The BPN method essentially runs both methods and gives
you a convenient way to compare fast NLP and accurate PAC results to see if
the NLP region of validity is acceptable for your simulation needs.
This section covers the following algorithms:
■
Nonlinear Perturbation Algorithm
■
Periodic AC Algorithm
■
Broadband Phase Noise Validation

HSPICE® User Guide: Advanced Analog Simulation and Analysis 101

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Nonlinear Perturbation Algorithm

The NLP algorithm, METHOD=0, has the fastest run time and gives very good
accuracy for core oscillator circuits and for circuits that do not include power
supply noise filtering.
The NLP approach performs a single frequency noise analysis about the
fundamental oscillation frequency, and then extrapolates this solution for offset
sweep frequencies based on the noise source behavior. This generally gives
very good results at low offset frequencies.
The NLP computation time has only a small dependence on the number of
frequency points in the phase noise frequency sweep.
Periodic AC Algorithm
The PAC algorithm, METHOD=1 (default), is intended for maximum accuracy
over a broad range of offset frequencies. Use the PAC approach to capture
accurate phase noise for oscillators that may have additional buffering, filtering,
dividers, and frequency dependent power supply noise effects.
The PAC approach performs a full analysis at each frequency given in the
phase noise frequency sweep. This makes it slower than the NLP algorithm,
but very accurate for capturing far-carrier noise filtering effects and measuring
the floor of the phase noise response.
The computation time for the PAC algorithm is approximately linearly
dependent on the number of frequency points in the phase noise sweep. So,
use a conservative number of sweep points for best performance. PAC
accuracy can generally be improved by using larger values for NHARMS (>16)
and RUNLVL (5 or 6). At low offset frequencies, the PAC approach will give
results identical to the NLP method.
Broadband Phase Noise Validation
The BPN validation, METHOD=2, gives a convenient method for comparing the
accuracy between the NLP and PAC approaches, over a broad offset frequency
range. BPN validation will run both the NLP and PAC algorithms and provide
data regarding their separate responses, their overlaps, and differences. The
resulting data and plots are ideal for identifying the offset frequency range of
validity of the NLP approach and showing where PAC accuracy is mandatory.
If the results reveal that the NLP algorithm has adequate accuracy, the future
simulation run-time can be much reduced by switching to the faster NLP
approach. Or the results may show that only the PAC method will give
acceptable results. The BPN validation results should be similar for oscillators
with similar topologies and noise contributions.

102 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

PHASENOISE Output Syntax

You can analyze element phase noise through the .PRINT and .PROBE
statements. HSPICE supports producing the output of the phase noise for the
whole circuit and the phase noise due to a specified element. In addition, by
using specialized keywords, you can output phase noise by noise source types.
To generate the output, you must enable (=on) the listsources option of the
.PHASENOISE command.
The following sections discuss these topics:
■
Whole Circuit and Specified Element Phase Noise
■
Frequency-Dependent and Frequency-independent Sources
■
Frequency and Bias Dependencies
■
Examples

Whole Circuit and Specified Element Phase Noise

A single phnoise keyword specifies the phase noise for the whole circuit and
the phnoise(element_name) specifies the phase noise value of a specified
element in the circuit.
.PRINT PHASENOISE phnoise phnoise(element_name)
.PROBE PHASENOISE phnoise phnoise(element_name)
In this syntax, the standalone phnoise is the phase noise parameter. For
example:
.PROBE PHASENOISE phnoise

The .PHASENOISE statement outputs raw data to the *.pn# and

*.printpn# files. HSPICE advanced analog analyses output the phnoise
data in decibels, relative to the carrier signal, per hertz, across the output
nodes in the .PHASENOISE statement (Units: dBc/Hz). The data plot is a
function of the offset frequency.
HSPICE advanced analog analyses output phnoise to the .pn# file if you
set .OPTION POST.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 103

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Frequency-Dependent and Frequency-independent Sources

■ The phnoise_fdep keyword variable collects all frequency-dependent
noise sources' contributions to the phase noise.
■
The phnoise_findep keyword variable collects all frequency
independent noise sources' contributions.
.print phasenoise phnoise_fdep
.print phasenoise phnoise_findep

104 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Frequency and Bias Dependencies

Table 1 Summary of Noise Type Dependences

Noise type frequency-dependent bias-dependent

phnoise_stationary No No

The following syntax is frequency-independent and bias-independent:

.print phasenoise phnoise_stationary

phnoise_cyclostationary or No Yes
phnoise_cyclo

The following syntax is frequency-independent and bias-dependent:

.print phasenoise phnoise_cyclo
or
.print phasenoise phnoise_cyclostationary
Where: cyclo or cyclostationary means anything bias-dependent.

phnoise_flicker Yes No

The following syntax is bias-independent and frequency-dependent:

.print phasenoise phnoise_flicker

phnoise_cycloflicker Yes Yes

The following syntax is frequency-dependent and bias-dependent:

.print phasenoise phnoise_cycloflicker
or
.print phasenoise phnoise_cyclostationaryflicker

phnoise_fdep is the union of phnoise_Flicker and

phnoise_cycloflicker noise types

phnoise_findep is the union of phnoise_stationary and

phnoise_cyclostationary noise types

Examples
Example 1
This example performs an oscillator analysis by searching for frequencies in
the vicinity of 900 MHz, followed by a phase noise analysis at frequency offsets
from 100 Hz to 10 MHz.
.HBOSC TONE=900MEG NHARMS=9
+ PROBENODE=gate,gnd,0.65

HSPICE® User Guide: Advanced Analog Simulation and Analysis 105

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

.PHASENOISE V(gate,gnd) DEC 10 100 1.0e7

+ METHOD=0 CARRIERINDEX=1 $use NLP algorithm
+ listsources=on
.PROBE PHASENOISE phnoise
.PRINT PHASENOISE phnoise(X1)

Example 2
This example performs a VCO analysis by searching for frequencies in the
vicinity of 2.4 GHz. This example uses eleven harmonics and sweeps the VCO
tuning voltage from 0 to 5 V. HSPICE advanced analog analyses use the
nonlinear perturbation (NLP) algorithm to perform a phase noise analysis
based on the fundamental frequency for each tuning voltage value.
.HBOSC TONE=2400MEG NHARMS=11
+ PROBENODE=drainP,drainN,1.0
+ FSPTS=20,2100MEG,2700MEG
+ SWEEP Vtune 0.0 5.0 0.2
.PHASENOISE V(drainP,drainN) DEC 10 100 1.0e7
+ METHOD=0 CARRIERINDEX=1 $use NLP algorithm
+ listsouces=on
.PROBE PHASENOISE phnoise
.PRINT PHASENOISE phnoise(X2)

See Also
■
Using Noise Analysis Results as Input Noise Sources

Measuring Phase Noise with .MEASURE PHASENOISE

The HSPICE advanced analog optimization flow can read the measured data
from a .MEASURE PHASENOISE analysis. You can combine this flow in the
HSPICE advanced analog analyses optimization routine with a .MEASURE
HBTR analysis (see Using .MEASURE with .HB Analyses) and a .MEASURE
HBNOISE analysis (see Measuring HBNOISE Analyses with .MEASURE).
The .MEASURE PHASENOISE syntax supports the following measurements:
■
FIND
.MEASURE PHASENOISE result FIND phnoise at = IFB_value
— yields the result of a variable value at a specific input frequency band
point. For example:
.MEASURE PHASENOISE np1 find PHNOISE at=100K
■
WHEN

106 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

.MEASURE PHASENOISE result WHEN phnoise=value

—yields the input frequency point at a specific phnoise value. For example:
.MEASURE PHASENOISE fcorn1 when PHNOISE=-120
■
RMS, average, min, max, and peak-to-peak
.MEASURE PHASENOISE result func phnoise
+ [FROM = IFB1] [TO = IFB2]
—yields the average, RMS, minimum, maximum, or peak-to-peak value of
the phase noise from frequency IFB1 to frequency IFB2, where the value
of func can be RMS, AVG, MIN, MAX or PP. If you do not specify FROM and
TO, HSPICE calculates the value over the frequency range that you specify
in the .PHASENOISE command. For example:
.measure PHASENOISE agn1 AVG phnoise from=100k to=10meg
■
Integral evaluation
.MEASURE PHASENOISE result INTEGRAL phnoise
+ [FROM = IFB1] [TO = IFB2]
—integrates the phase noise value from the IFB1 frequency to the IFB2
frequency. For example:
.MEASURE PHASENOISE rns1 INTEGRAL phnoise from=50k to 500k
■
Derivative evaluation
.MEASURE PHASENOISE result DERIVATIVE phnoise AT = IFB1
—finds the derivative of phase noise at the IFB1 frequency point. For
example:
.MEASURE PHASENOISE fdn1 DERIVATIVE phnoise at=10meg

Note: .MEASURE PHASENOISE cannot contain an expression that uses

a phase noise variable as an argument. You also cannot
use .MEASURE PHASENOISE for error measurement and
expression evaluation of the .PHASENOISE command.

See also the .MEASURE PHASENOISE command in the HSPICE Reference

Manual: Commands and Control Options.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 107

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Optimizing with PHASENOISE Measurement

The required statements are:
■
Analysis statement
.HB TONES=f1[f2...fn] [NHARMS=h1 [,h2...hn]]
+ SWEEP OPTIMIZE=OPTxxx RESULT=measname MODEL=mname

For example:
.HBOSC tone=1g nharms=5 PROBENODE=out,gnd,0.8
+ SWEEP OPTIMIZE=opt1 RESULT=y1,y2 MODEL=m1
.MODEL m1 OPT level=0
.PHASENOISE v(out) DEC 1 1k 1G
.MEAS PHASENOISE y1 FIND PHNOISE at 10k goal=-150dBc
.MEAS PHASENOISE y2 RMSJITTER PHNOISE units=sec goal=1p
■
Measure statement
.MEASURE PHASENOISE measname FIND out_var1 AT=val
+ GOAL=val

Amplitude Modulation/Phase Modulation Separation

You can separate the Amplitude Modulation (AM) and Phase Modulation (PM)
components of the total noise by calculating components in-phase (AM
component) and in quadrature (PM component) with the carrier, by using PAC
and BPN PHASENOISE analysis. The output and measure syntax separates
AM/PM noise.
Turn this feature on by setting the .OPTION PHNOISEAMPM=1 (see .OPTION
PHNOISEAMPM in the HSPICE Reference Manual: Commands and Control
Options. See also, Important Note for AM/PM Users at the end of this section).
■
If you use the NLP algorithm (METHOD=0), HSPICE advanced analog
analyses calculate only the phase noise component.
■
If you use either the PAC algorithm (METHOD=1) or the BPN algorithm
(METHOD=2), HSPICE advanced analog analyses add both the phase and
amplitude noise components together to show the total noise at the output.
The following sections discuss these topics:
■
AM/PM .PRINT and .PROBE Statement Syntax
■
Examples

108 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

■
Output Files
■
Important Note for AM/PM Users

AM/PM .PRINT and .PROBE Statement Syntax

.PROBE PHASENOISE phnoise [la] [ltotal] [onoise]
.PRINT PHASENOISE phnoise [la] [ltotal] [onoise]

Keywords for AM/PM separations are:

■
Phase Modulation term only: phnoise
■
Amplitude Modulation term only: la
■
Total phase noise term: ltotal
■
Voltage noise term: onoise
For example:
.probe phasenoise phnoise la ltotal $ AM modulation and
total phase noise

Add a Noise type suffix to each of these noise terms, phnoise, la, ltotal,
onoise, to select specific noise-type components:

Table 2 Summary of Noise_term

Noise type frequency-dependent bias-dependent

Noise_term_phnoise_stationary No No

Noise_term_phnoise_cyclostationary No Yes

Noise_term_phnoise_flicker Yes No

Noise_term_phnoise_cycloflicker Yes Yes

Noise_term_phnoise_fdep The union of phnoise_Flicker and

phnoise_cycloflicker noise types

Noise_term_phnoise_findep The union of phnoise_stationary and

phnoise_cyclostationary noise types

Examples
Example 1
.PROBE PHASENOISE la_phnoise_cyclostationary

HSPICE® User Guide: Advanced Analog Simulation and Analysis 109

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

You can also show the AM/PM separation for individual noise elements. (To
enable this capability, the listsources option of .PHASENOISE must = on.)
This example outputs the phase modulation noise associated only with cyclo-
stationary sources (that is, sources that are bias dependent, but not frequency
dependent).
Noise Element output is of the form Noise_term(element_name), where
Noise_term can be phnoise, la, ltotal, or onoise, and element_name is
a valid netlist element name.
Example 2
.PROBE PHASENOISE la(x1)

Output Files
■
File *.printpn#: Writes output from the .PRINT statement when using
HB to obtain the steady state solution.
■
File *.pn#: Writes output from the .PROBE statement when using HB to
obtain the steady state solution.
■
File *.printsnpn#: Writes output from the .PRINT statement when using
SN to obtain the steady state solution.
■
File *.snpn#: Writes output from the .PROBE statement when using SN to
obtain the steady state solution.
The .PHASENOISE command line parameters Listfreq, ListCount,
Listfloor, and Listsources control and list noise source contributions
sequentially.The listsources argument must be =on to generate a noise list
block for each output parameter specified in the .PRINT/.PROBE statement
such as phnoise, la, ltotal, and onoise.
.MEASURE Syntax and Outputs
.MEASURE PHASENOISE extends output variables to the set: am[noise]
pm[noise]
.MEASURE output files include:
■
File *.mpn#: Writes output from the .MEASURE statement when using HB
to obtain the steady state solution.
■
File *.msnpn#: Writes output from the .MEASURE statement when using
SN to obtain the steady state solution.

110 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Important Note for AM/PM Users

There are discrepancies that may occur between this feature and the traditional
PAC phase noise analysis in HSPICE advanced analog analyses. Total phase
noise (ltotal) is the sum of two terms, the amplitude modulation (am) and
phase modulation (phnoise). Traditionally, PAC phase noise reports the
phnoise (phase modulation component) and ltotal (total phase noise)
terms as identical, with the assumption that the am term (amplitude modulation
component) is zero.
The PAC phase noise AM/PM feature described in this section separately
calculates the am and phnoise components. The AM/PM feature affects all
phase noise measurements that involve either PAC or BPN. In most cases the
differences between PHNOISEAMPM=1 and =0 are small, unless you expect a
significant AM component. You may see a slight decrease in the new phase
noise (phase modulation) component when compared to the old calculation.
For example, the random jitter calculations are accurate only when they involve
the PM component of phase noise. When basing the calculation on ltotal, or
if you include the AM noise, the process may introduce a small error.
NLP phase noise (method=0) only calculates the phnoise component; the
AM/PM option does not affect this method.
BPN phase noise (method=2) is affected in that the far side component is
derived from the PAC phase noise.
Workaround: Ensure that.OPTION PHNOISEAMPM=0 (the default). This
assures that periodic AC phase noise amplitude-modulation (AM) component is
zero to maintain backward compatibility and produce the traditional results for
phnoise and jitter.

Interpreting Phase Noise Analysis Results

A typical phase noise plot consists of a line which drops off as a function of
frequency, at a slope of -20dbc/decade where white noise dominates, or -
30dbc/decade where flicker noise dominates. At very low offset frequencies,
the phase noise rolls off according to a Lorentzian shape, such that it never
exceeds 0 dbc/Hz even for very low offset frequencies. The 0 dbc/Hz value
represents the power of the carrier oscillation, at 0 offset frequency. At very
high offset frequencies, the slope can deviate from -20 dbc/decade due to the
existence of a noise floor or a circuit feedback effect.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 111

N-2017.12
Chapter 5: Phase Noise Analysis
Introduction to Phase Noise Analysis (.PHASENOISE)

Numerical methods for phase noise analysis have limitations. The main
limitation in the PAC phase noise algorithm is that it rolls off too quickly at low
offset frequencies. In the low frequency region, you can trust NLP. The main
limitation of the NLP algorithm is that it does not cover all high frequency
effects, so you can trust PAC in the high frequency region.
The BPN algorithm attempts to combine the NLP and PAC results to generate a
single result that is valid for all offset frequencies. It may fail if it cannot identify
an overlap region where NLP and PAC results match. If the tool cannot find an
overlap region, you should attempt to increase nharms in the .HBOSC
command, as this increases the accuracy of both algorithms, especially PAC.
PAC accuracy is more sensitive to nharms than NLP.
If you suspect the phase noise results to be inaccurate, check the following:
1. Is the .HBOSC steady state solution fully converged?
Explanation: The NLP or PAC small-signal noise analysis requires a highly
accurate steady state solution.
2. Did the phase noise analysis fully converge?
Explanation: Phase noise analysis uses a GMRES iterative linear solver.
If this iterative solver reaches its iteration limit before full convergence, the
results are not reliable. Check the number of Krylov iterations that the phase
noise analysis required. If it took the maximum number of iterations (as set
by PHASENOISE_KRYLOV_ITR, default=1000), then the results did not fully
converge and you should not trust them.
You can use the options PHASENOISE_KRYLOV_DIM and
PHASENOISE_KRYLOV_ITR to control the GMRES solver. You can
increase PHASENOISE_KRYLOV_DIM to improve the convergence rate at
the expense of memory, or increase PHASENOISE_KRYLOV_ITR to allow
more iterations.
3. Is the phase noise analysis accurate?
Explanation: In rare circumstances, the linear solver accuracy needs to be
improved. The linear solver accuracy is controlled by the PHASENOISETOL
option. The default value of the option is 1e-9. The suggested setting for
PHASENOISETOL is 1e-10, if the accuracy needs to be improved.

112 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

Accumulated Jitter Measurement for Closed Loop PLL

Analysis
Enhancements to HSPICE advanced analog analyses include considerable
support for a variety of jitter measurements. Many of these are important in a
PLL flow, where you use the HBOSC or SNOSC analyses to compute a phase
noise response for an oscillator or VCO, to derive the resulting random jitter
from phase noise. In the PLL methodology, you use other HSPICE advanced
analog analyses to compute phase noise contributions from the other PLL
building blocks. You then perform a closed loop analysis by using phase-
domain models for both signal and noise responses that take into account the
noise contributions from all such blocks. To complete this flow is the ability to
compute “Accumulated Jitter” or “Timing Jitter” for the closed loop PLL. The
accumulated jitter response is essentially an integral transformation of the
closed-loop PLL response. The following sections show how you can measure
accumulated jitter directly from the phase noise-output of an open loop
oscillator analysis.
In the PLL flow, you interpret the closed loop phase noise from the results of a
linear HSPICE advanced analog .AC/.NOISE analysis. The following sections
describe a capability that allows direct measurement of accumulated jitter from
the results of this closed loop noise analysis, without any special interpretation
of the results.
This section covers the following topics:
■
Jitter Measurements from Phase Noise

Jitter Measurements from Phase Noise

This section discusses the following topics:
■ Jitter Definitions
■
Jitter Output Syntax
■
.MEASURE Statements for Jitter
■
Peak-to-Peak Jitter
■
Errors/Warnings

HSPICE® User Guide: Advanced Analog Simulation and Analysis 113

N-2017.12
Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

Jitter Definitions
HSPICE advanced analog analyses provide several random jitter (RJ)
measurements. This section defines, describes, and compares the various jitter
measurements provided. You derive random jitter measurements from the
results of an HSPICE advanced analog phase noise analysis. The following
presents the relationships between phase noise and the random jitter
measurements, and their means for calculation. The types of random jitter
measurements include: Timing, Phase, Period, Tracking, Long-Term, and
Cycle-to-Cycle Jitter.
Timing Jitter is a measurement of the oscillator uncertainty in the time domain.
For clock applications, time domain measurements are preferable, since most
specifications of concern involve time domain values.
Timing Jitter is the standard deviation of the timing uncertainty, which is a
function of the auto-correlation function in the power spectrum of the phase
variations. Timing jitter is the square root of the variance (standard deviation
squared) of the timing uncertainty between two clock edges separated by an
interval given by  = N  T o , where T o is the ideal clock period. You can write it
as a function of the auto-correlation function of the power spectrum of phase
variations as:

Equation 24
2 2-
 TIE    = --------
2
 R  0  – R    
o
where TIE refers to the Time Interval Error. Call this measurement Timing
Jitter, Accumulated Jitter, or N-Cycle Jitter, since it represents the jitter that
may accumulate over an interval of many periods.
The Weiner-Khintchine Theorem [1] relates the auto correlation function to the
power spectrum of phase variations as in the following equation:

Equation 25
 
1
 S   e d = 2  L  f  cos  2f  df
j
R     = ------
2
– 0

where S     is the double-sided power spectrum of phase variations, and L  f 

is the single-sideband phase noise. The auto-correlation for  = 0 is given by:

114 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

Equation 26


= 2  L  f  df
2
R  0    rms
0

which defines  rms in HSPICE advanced analog analyses known as RMS

Phase Jitter.
2
Using the identity 2 sin  = 1 – cos 2 we can then write:

Equation 27

8
2 
2 2
 TIE    = --------
- L  f  sin  f  df
o
0
to enable currently supported HSPICE advanced analog analyses jitter
measurements to be written as:

Equation 28

 rms =  ph   0 = 2  L  f  df “RMS Phase Jitter”

0

2
 TIE    = ------ 2  L  f sin  f df “Timing (Time Interval Error) Jitter”
2
0
0

From these definitions, several other key jitter measurements can be derived,
including Period Jitter, Tracking Jitter, Long-Term Jitter, and Cycle-to-Cycle
Jitter.
Period Jitter is equivalent to the value for Timing Jitter for a one period interval.
We therefore have:

Equation 29

2
=  TIE  T 0  = ------ 2  L  f sin  fT 0 df
2
 PER "Period Jitter"
0
0
Tracking Jitter is equivalent to the value (in units of seconds) for RMS Phase
Jitter, or:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 115

N-2017.12
Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

Equation 30

 rms 1
 tr =  ph = ---------- = ------ 2  L  f df “Tracking Jitter”
0 0
0

Long-Term Jitter is equivalent to 2 times the Tracking Jitter, that is:

Equation 31

 rms 2
 T   =  TIE      = 2 ---------- = ------
0 0 L  f  df “Long-Term Jitter”
0
Cycle-to-Cycle Jitter is based on the difference between adjacent Period Jitter
measurements. It is given by:

Equation 32
2 2
 CTC = 4PER –  TIE  2T 0  “Cycle-to-Cycle Jitter”
In general, each of the above calculations must be performed carefully over
limits of integration to accurately calculate jitter expressions based on the finite
frequency limits provided for the phase noise analysis. Linear interpolation is
used, but the phase noise generally follows more of a power law expansion.

Jitter Output Syntax

The timing jitter calculations are derived from the results of phase noise
analysis. The phase noise output syntax supports the JITTER keyword as an
output keyword in addition to the PHNOISE keyword.
.PRINT PHASENOISE PHNOISE JITTER
.PROBE PHASENOISE PHNOISE JITTER
If the JITTER keyword is present, the .PHASENOISE statement also outputs
the raw jitter data to *.jt0 and *.printjt0 data files. The PHNOISE data is
given in units of dBc/Hz, that is, dB relative to the carrier, per Hz, across the
output nodes specified by the PHASENOISE statement. The data plot is a
function of offset frequency. If the JITTER keyword is present, .PHASENOISE
outputs the TIE Timing Jitter (Accumulated Jitter) data to *.jt0 and
*.printjt0 data files. These data are plotted as a function of time in units of
seconds. The jitter calculations make use of some of the parameters given in
the .PHASENOISE syntax.

116 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

The time samples for the timing jitter output make use of the same number of
points as the phase noise frequency sweep specification.
The output of the timing jitter information uses a corresponding time sampling
derived via:

Equation 33
1 2 N
 1 = ------  2 = ------ ..., N = ------
T0 T0 T0

.MEASURE Statements for Jitter

The jitter-specific .MEASURE statements specify the jitter keywords as follows.
(For a discussion of the BER parameter, see below.)
.MEASURE PHASENOISE Jname PERJITTER phnoise
+ [UNITS=(sec|rad|UI)] [BER=val]
.MEASURE PHASENOISE Jname CTCJITTER phnoise
+ [UNITS=(sec|rad|UI)] [BER=val]
.MEASURE PHASENOISE Jname RMSJITTER phnoise
+ [FROM start_frequency] [TO end_frequency]
+ [UNITS=(sec|rad|UI)] [BER=val]
.MEASURE PHASENOISE Jname PHJITTER phnoise
+ [FROM start_frequency] [TO end_frequency]
+ [UNITS=(sec|rad|UI)] [BER=val]
.MEASURE PHASENOISE Jname TRJITTER phnoise
+ [FROM start_frequency] [TO end_frequency]
+ [UNITS=(sec|rad|UI)] [BER=val]
.MEASURE PHASENOISE Jname LTJITTER phnoise
+ [FROM start_frequency] [TO end_frequency]
+ [UNITS=(sec|rad|UI)] [BER=val]
RMSJITTER, PHJITTER, and TRJITTER are synonymous measurements, all
based on the calculations described related to the RMS Phase Jitter value in
units of seconds given by  ph =  rms   0 . These measurements allow control
of the integration range using the FROM and TO parameters. The
measurements for PERJITTER and CTCJITTER use the full offset frequency
sweep range given for the phase noise analysis to compute values (the FROM
and TO parameters are ignored if entered).
In HSPICE advanced analog analyses, the frequency intervals can be modified
for these jitter calculations (if desired, although not recommended), and UNITS
can be selected between seconds, radians, and Unit Intervals. The following

HSPICE® User Guide: Advanced Analog Simulation and Analysis 117

N-2017.12
Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

table specifies the calculation used for units=seconds for each jitter
measurement.

MEASURE name Calculation used (Units=sec)

RMSJITTER  ph =  rms   0

PHJITTER  ph =  rms   0

TRJITTER  ph =  rms   0

PERJITTER  PER

LTJITTER
 T   = 2 rms   o

CTCJITTER  CTC

Example:
.meas phasenoise rj RMSJITTER phnoise from 1K to 100K
+ units = rad

Peak-to-Peak Jitter
As noted in .MEASURE Statements for Jitter, an additional Bit Error Rate (BER)
parameter is supported. This parameter allows you to convert any jitter value
from an RMS value into a Peak-to-Peak value. The RMS jitter values
correspond to an 1-sigma standard deviation value for the Gaussian distribution
of the jitter. Peak-to-peak values represent the full span of the Gaussian
distribution. Since this span is theoretically unbounded for truly random
distributions, the conversion to peak-to-peak values has to be interpreted as
spanning some number of sigma values. You can arrive at this number (in other
words, “sigma multiplier”) by specifying a corresponding BER.
The term “BER” corresponds to the unitless Bit Error Rate that allows for this
conversion. The following table shows some sample conversions from various

118 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 5: Phase Noise Analysis
Accumulated Jitter Measurement for Closed Loop PLL Analysis

BER values into a “sigma multiplier” value, which corresponds to the number of
sigma standard deviations in converting from RMS to peak-to-peak values:

Bit Error Rate Sigma Multiplier

10-3 6.180

10-4 7.438

10-5 8.530

10-6 9.507

10-7 10.399

10-8 11.224

10-9 11.996

10-10 12.723

10-11 13.412

10-12 14.069

10-13 14.698

10-14 15.301

10-15 15.883

10-16 16.444

These conversions are done in accordance with the relationship:

Equation 34
1--- 
erfc -------------- = BER
2 2 2
where, erfc is the complementary error function and  is the Sigma Multiplier.
Support for peak-to-peak conversions is included for a continuous range of
– 16 –3
BER values from 10    10 (and some values extrapolated outside
this range).

HSPICE® User Guide: Advanced Analog Simulation and Analysis 119

N-2017.12
Chapter 5: Phase Noise Analysis
Clock Source with Random Jitter

Specification of the BER parameter results in the output of the peak-to-peak

jitter value and not the RMS value. Labels for the measurements show
appropriate “rms” and “p-p” labels. A BER parameter set to BER=0 is equivalent
to having no parameter and only results in the RMS calculation.

Errors/Warnings
Error handling and recovery is exercised to capture obvious errors in input
specifications. The following error checks are performed:
■
Calculations to be performed if oscillator or phase noise analysis fails.
■
ERROR if L(f) > 1 over any part of the frequency sweep (non-dB form).
■
ERROR if L(f) < 0 over any part of the frequency sweep (non-dB form).
■
Error if any time or frequency samples are negative values.
■
ERROR if BER < 0 for any Jitter measurement.
■
WARNING if BER > 1 for any Jitter measurement.
■
WARNING if f0 < 10 Hz. Message: “Jitter calculations may be
ineffective for offset frequencies under 10 Hz.”

Clock Source with Random Jitter

For .AC-related .NOISE analysis, see Clock Source with Random Jitter in the
HSPICE® User Guide: Elements.

120 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 6
6 Behavioral PLL Noise Analysis
(.ACPHASENOISE)

Describes behavioral PLL noise analysis.

To see the influence that oscillator or VCO phase noise can have on a system
where it is present, it is necessary to perform a phase-domain analysis where
the circuit variables are phase and the input noise stimuli are phase noise. This
is the purpose of the .ACPHASENOISE analysis in HSPICE advanced analog
analyses.
This type of analysis is critical, for example, in analyzing the effects of noise in
a phase-locked loop (PLL). In a PLL design flow, the HBOSC or SNOSC
analyses are used to compute a phase noise response for an oscillator or VCO.
HSPICE advanced analog analyses can be used to compute phase noise
contributions from the other PLL building blocks. A closed loop PLL analysis
can then be performed by using phase-domain models for both signal and
noise responses, where the noise contributions from all blocks are input as
phase noise stimuli. Such an analysis can be performed to determine the PLL
closed-loop phase noise, based on the contributions of each block, determined
by an open loop analysis.
For more information on control options, see the .ACPHASENOISE command
in the HSPICE Reference Manual: Commands and Control Options.
This chapter discusses the following topics:
■
ACPHASENOISE Analysis .PRINT and .PROBE Syntax
■ Behavioral Noise Sources

HSPICE® User Guide: Advanced Analog Simulation and Analysis 121

N-2017.12
Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
ACPHASENOISE Analysis .PRINT and .PROBE Syntax

ACPHASENOISE Analysis .PRINT and .PROBE Syntax

The unique aspect of the .ACPHASENOISE analysis is that it allows the small
signal noise calculation results to be interpreted as phase noise values. The
available .print/.probe measurements reflect this. The .print/.probe
output syntax are the “JITTER” and “PHNOISE” keywords consistent with the
HSPICE advanced analog .phasenoise analysis, namely:
.PRINT ACPHASENOISE PHNOISE JITTER
.PROBE ACPHASENOISE PHNOISE JITTER
As with the .PHASENOISE analysis, the .ACPHASENOISE analysis outputs raw
data to *.pn0 and *.printpn0 files. The PHNOISE data is given in units of
dBc/Hz, that is, dB relative to the carrier, per Hz, across the output nodes
specified by the .ACPHASENOISE statement. The data plot is a function of
offset frequency. If the “JITTER” keyword is present, .ACPHASENOISE also
outputs the accumulated TIE jitter data to *.jt0 and *.printjt0 data files.
These data are plotted as a function of time in units of seconds. The Timing
Jitter data itself has units of seconds. The timing jitter calculations make use of
the parameters given in the .ACPHASENOISE syntax, such as “freq” and
“interval”.
See also Using Noise Analysis Results as Input Noise Sources.

.MEASURE Support for ACPHASENOISE

Single-valued jitter measurements are available from .MEASURE statements.
Examples include period jitter, cycle-to-cycle jitter, and phase jitter
measurements, as shown respectively in the following syntax:
.MEASURE ACPHASENOISE Jname PERJITTER phnoise
+ [UNITS=(sec|rad|UI)] [BER=val]

.MEASURE ACPHASENOISE Jname CTCJITTER phnoise

+ [UNITS=(sec|rad|UI)] [BER=val

.MEASURE ACPHASENOISE Jname PHJITTER phnoise

+ [FROM start_frequency [TO end_frequency]
+ [UNITS=(sec|rad|UI)] [BER=val]

122 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

Behavioral Noise Sources

In HSPICE advanced analog analyses, you can use the G-element to specify
noise sources. Frequency domain noise analyses (.NOISE, .HBNOISE, and
.PHASENOISE) take these noise sources into account.
You can attach noise sources to behavioral models. For example, you can use
a G-element with the VCCAP parameter to model a varactor, which includes a
noise model. You can also simulate effects such as substrate noise, including
its effect on oscillator phase noise. You can also use this G-element syntax to
simulate behavioral descriptions of substrate noise during any frequency
domain noise analysis, which includes phase noise analysis. For example,
gname node1 node2 noise=’noise_equation’
gname node1 node2 node3 node4 noise=’noise_equation’

The first line creates a simple two-terminal current noise source, whose value
is described in A2/(Hz). The output noise generated from this noise source is:
noise_equation*H
Where H is the transfer function from the terminal pair (node1,node2) to the
circuit output, where HSPICE advanced analog analyses measures the output
noise.
The second line produces a noise source correlation between the
(node1,node2) and (node3,node4) terminal pairs. The resulting output noise
is calculated as noise_equation*sqrt(H1*H2*); where:
■
H1 is the transfer function from (node1,node2) to the output.
■ H2 is the transfer function from (node3,node4) to the output.
■
The * on H1*H2* represents the complex conjugate of H1 and H2.
The noise_equation expression can involve node voltages and currents
through voltage sources.
For the PAC phase noise simulation to evaluate the frequency-dependent
noise, the frequency-dependent noise factor in the phase noise must be
expressed between parentheses. For example:
gname node1 node2 noise = '(frequency_dependent_noise)*
bias_dependent_noise'

This is only true when the total noise can be expressed in this form and when
the frequency-dependent noise can be evaluated in the PAC phase noise

HSPICE® User Guide: Advanced Analog Simulation and Analysis 123

N-2017.12
Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

simulation. You can also input the behavioral noise source as a noise table with
the help of the predefined Table() function. The Table() function takes two
formats:
■
Noise table can be input directly through the Table() function. For
example:
gname node1 node2 noise = 'Table(arg1,f1,v1,f2,v2,......)'
■
The f1,v1,f2,v2,..... parameters describe the noise table. When arg1 ==
f1, the function returns v1. The arg1 can be an expression of either
HERTZ, bias, or both. For example, arg1 = 'HERTZ * 1.0E+3'.
■
The noise table can be input through a .DATA structure:
.DATA d1
+ x y
+ f1 v1
+ f2 v2
.ENDDATA

gname node1 node2 noise = 'TABLE(arg1,d1)'

The x, y parameters in the DATA structure are two placeholder strings that can
be set to whatever you prefer even if they are in conflict with other parameters
in the netlist. The arg1 parameter can be an expression of HERTZ and bias.
When arg1 == f2, the function returns v2.
This section covers the following topics:
■
Using Noise Analysis Results as Input Noise Sources
■
Syntax Rules for Behavioral Noise Source Files
■
Power Supply Current and Voltage Noise Sources

Using Noise Analysis Results as Input Noise Sources

By extending the E- and G- voltage-controlled source syntax, the phase noise
or noise data in the ASCII phase noise and noise files can used as input for
specifying behavioral noise sources.
Usage Model
The syntax for the voltage-controlled voltage (E) or current (G) source is:
Exxx node1 node2 noisefile='filename' [mname='measname']
Gxxx node1 node2 noisefile='filename' [mname='measname']

124 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

where:
noisefile='filename' is the name of the ASCII noise data file. Phase
noise analysis and noise analyses can output simulation results as ASCII data
using .PRINT statements. The file name is designated as
'design.printsnpn0' for .SNOSC phase noise and .SNNOISE analyses.
For .PHASENOISE, .HBNOISE and .ACPHASENOISE analyses, the file name
is designated as 'design.printpn0'. For .NOISE analyses, the file name
is designated as 'design.printac0'.
mname='measname' is used to select the appropriate noise measurement
name to be taken from the *.printpn0, *.printsnpn0, or *.printac0
file.
measname can be one of the following:
■
NLP_L(f) – selects the nlp_L(f) phase noise data in units of dBc/Hz.
■
PAC_L(f) – selects the pac_l(f) phase noise data in units of dBc/Hz.
■
BPN_L(f) – selects the bpn_l(f) phase noise data in units of dBc/Hz.
■
ONOISE – selects the onoise data based on .SNNOISE or .NOISE analysis.
The following syntaxes are supported in both HSPICE advanced analog
analyses and HSPICE:
■
Exxx n1 n2 noise data=dataname
■
Exxx n1 n2 noise data=datablock
■
Exxx n1 n2 noisefile='filename'
■
Exxx n1 n2 noise='expression'
■
Exxx n1 n2 noise='Table(arg1,f1,v1,f2,v2...)'
■ Exxx n1 n2 noise='Table(arg1,dotDataBlockName)', where
dotDataBlockName is the .data statement reference.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 125

N-2017.12
Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

Syntax Rules for Behavioral Noise Source Files

You can use ASCII noise data files not generated by .PRINT statements when
you follow these syntax rules:
■
A noise file always needs a header (do not include frequency or noise points
in the header line, as they will be ignored).
■
The items in the header must be consistent with the number of data
columns.
■
You can use noise keywords in the header (see the next item in this list).
■
The noise value units depend on your use of the MNAME keyword.
You can use noise keywords like ONOISE, and the phase noise keywords,
PAC_L(f), BPN_L(f), and, NLP_L(f). This is useful for *.printac#,
*.printpn#, or *.printsnpn# files and use of the MNAME keyword to
specify the noise data column to be used. For example, if you do a phase noise
analysis and specify broadb and phase noise (method=2 in the .PHASENOISE
command), the *.printpn# file contains three columns of phase noise data:
PAC_L(f), BPN_L(f), and, NLP_L(f).
If MNAME is not specified with the behavioral noise source, then HSPICE
assumes that the second column in the noise file represents the desired noise
values, and the MNAME string is read from the noise file header. The first
column is always assumed to contain frequency values.
The MNAME keyword is used to differentiate units. If you use MNAME=ONOISE
and the file contains ONOISE in the header, then the noise units are interpreted
as V/sqrt(Hz). The same is true with MNAME=ONOISE(M) and
MNAME=ONOISE(MAG). If MNAME is set to PAC_L(f), BPN_L(f), NLP_L(f),
or PHNOISE, then units are interpreted as dBc/Hz. With MNAME=ONOISE(DB),
units are taken as dB/Hz (20*Log(onoise)), as are other MNAME keywords that
end in DB or (DB). Most other MNAME keywords are generally interpreted in
V^2/Hz units, unless the name ends in(M) or (MAG), which are taken to be
V/sqrt(Hz). Depending on the noise units, different interpolation schemes are
used for frequencies needed between noise data points. Noise in dB/Hz or
dBc/Hz units use logarithmic interpolation (in x and y), while non-dB units use
linear interpolation in V^2/Hz.

Noise File Examples

The following examples provide sample noise results noting the various noise
value units:

126 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

Example 1: Phase noise file generated by using .PRINT phasenoise

phnoise statement. The noise units are dBc/Hz:
HERTZ NLP_L(f)
10.0000000000k -130.2612492663
12.5892541179k -132.2612484153
15.8489319246k -134.2612470666
19.9526231497k -136.2612449295
25.1188643151k -138.2612415429
31.6227766017k -140.2612361768
39.8107170553k -142.2612276743
50.1187233627k -144.2612142034
63.0957344480k -146.2611928626
79.4328234724k -148.2611590575
100.0000000000k -150.2611055150
125.8925411794k -152.2610207237
158.4893192461k -154.2608864692
199.5262314969k -156.2606739394
251.1886431510k -158.2603375701
316.2277660168k -160.2598053187
398.1071705535k -162.2589632779
501.1872336273k -164.2576312765
630.9573444802k -166.2555240055
794.3282347243k -168.2521885144
1.0000000000x -170.2469020563

Example 2: Noise file generated by using .PRINT snnoise onoise

statement. The noise units are V/sqrt(Hz):
HERTZ ONOISE()
100.0000 1.6699u
526.4105k 46.8034n
1.0527x 45.1538n
1.5790x 44.6096n
2.1053x 44.3405n
2.6317x 44.1804n
3.1580x 44.0744n
3.6843x 43.9990n
4.2106x 43.9425n
4.7369x 43.8986n
5.2632x 43.8634n
5.7895x 43.8345n
6.3158x 43.8102n
6.8421x 43.7894n
7.3684x 43.7714n
7.8948x 43.7556n
8.4211x 43.7415n
8.9474x 43.7288n
9.4737x 43.7172n

HSPICE® User Guide: Advanced Analog Simulation and Analysis 127

N-2017.12
Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

10.0000x 43.7067n

Example 3: User-generated noise file. Units are in dBc/Hz:

HERTZ NOISE(DB)
10.0 -120
100.0 -125
1.0k -130
5.0k -135
10.0k -140
100.0k -150

Power Supply Current and Voltage Noise Sources

You can implement the power supply noise source with G- and E- elements.
The G-element is for the current noise source and the E-element for the voltage
noise source. As noise elements, they are two-terminal elements that represent
a noise source connected between two specified nodes.
■
Syntax
■
Example

Syntax
Expression form:
Gxxx node1 node2 noise=‘expression’
Exxx node1 node2 noise=‘expression’
The G noise element represents a noise current source and the E noise
element represents a noise voltage source. The xxx parameter can be set with
a value up to 1024 characters. The node1 and node2 are the positive and
negative nodes that connect to the noise source. The noise expression can
contain the bias, frequency, or other parameters.
Data form:
Gxxx node1 node2 noise data=dataname
Exxx node1 node2 noise data=dataname
.data dataname
+ pname1 pname2
+ freq1 noise1
+ freq2 noise2
+ ...
.enddata

128 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

The data form defines a basic frequency-noise table. The .DATA statement
contains two parameters: frequency and noise to specify the noise value at
each frequency point. The unit for frequency is hertz, and the unit for noise is
A2/Hz (for G current noise source) or V2/Hz (for E voltage noise source).

Example
The following netlist shows a 1000 ohm resistor (g1) using a G-element. The
g1noise element, placed in parallel with the g1 resistor, delivers the thermal
noise expected from a resistor. The r1 resistor is included for comparison. The
noise due to r1 should be the same as the noise due to g1noise.
* Resistor implemented using g-element
v1 1 0 1
r1 1 2 1k
g1 1 2 cur='v(1,2)*0.001'
g1noise 1 2
+ noise='4*1.3806266e-23*(TEMPER+273.15)*0.001'
rout 2 0 1meg
.ac lin 1 100 100
.noise v(2) v1 1
.end

HSPICE® User Guide: Advanced Analog Simulation and Analysis 129

N-2017.12
Chapter 6: Behavioral PLL Noise Analysis (.ACPHASENOISE)
Behavioral Noise Sources

130 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Part 2: Periodically-driven AC,
Transfer Function, and Noise
Analyses

The following chapters/topics are included in this Part:

■ Chapter 7, Large Signal AC, Transfer Function, and Noise Anal-
yses
■
Chapter 8, Large Signal S-parameter Analyses
■ Chapter 9, Envelope Analysis

HSPICE® User Guide: Advanced Analog Simulation and Analysis 131

N-2017.12
132 HSPICE® User Guide: Advanced Analog Simulation and Analysis
N-2017.12
 7
7 Large Signal AC, Transfer Function, and Noise
Analyses

Describes how to use both harmonic balance-based and Shooting Newton-

based AC and transfer function analyses, as well as nonlinear, steady-state
noise analysis.

This chapter discusses the following topics:

■
Multitone Harmonic Balance AC Analysis (.HBAC)
■ Shooting Newton AC Analysis (.SNAC)
■
Steady-State Voltage and Current Sources
■
Phase Noise and Buffer Chains
■ Multitone Harmonic Balance Noise (.HBNOISE)
■
Shooting Newton Noise Analysis (.SNNOISE)
■
Multitone Harmonic Balance Transfer Function Analysis (.HBXF)
■ Shooting Newton Transfer Function Analysis (.SNXF)
■
References

Multitone Harmonic Balance AC Analysis (.HBAC)

You use the .HBAC (Harmonic Balance AC) statement for analyzing linear
behavior in large-signal periodic systems. The .HBAC statement uses a
periodic AC (PAC) algorithm to perform linear analysis of autonomous
(oscillator) or non-autonomous (driven) circuits, where the linear coefficients
are modulated by a periodic, steady-state signal.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 133

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

Multitone HBAC analysis extends single-tone HBAC to quasi-periodic systems

with more than one periodic, steady-state tone. One application of multitone
HBAC is to more efficiently determine mixer conversion gain under the
influence of a strong interfering signal than is possible by running a swept
three-tone HB simulation.
The following sections discuss these topics:
■
Prerequisites and Limitations
■
Tutorial Example - Using Multi-Tone HB and HBAC Analyses for a Mixer
■
Input Syntax
■
Output Syntax
■
HBAC Output Data Files
■
Using the .MEASURE Command with .HBAC
■
Errors and Warnings

Prerequisites and Limitations

The following prerequisites and limitations apply to HBAC:
■
Requires one and only one .HBAC statement. If you use multiple .HBAC
statements, HSPICE advanced analog analyses use only the last .HBAC
statement.
■
Requires one and only one .HB statement.
■
Supports arbitrary number of tones.
■
Requires placing the parameter sweep in the .HB statement.
■
Requires at least one HB source.
■
Requires at least one HBAC source.
■
Supports unlimited number of HB and HBAC sources.
■
The requested maximum harmonic in a .PROBE or .PRINT statement must
be less than or equal to half the number of harmonics specified in harmonic
balance (that is, max_harm <= num_hb_harms / 2).

134 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

Tutorial Example - Using Multi-Tone HB and HBAC

Analyses for a Mixer
The example in this section shows how to use HSPICE advanced analog
analyses to analyze a circuit driven by multiple input stimuli with different
frequencies. Mixer circuits provide a typical example of this scenario: in this
case, there might be two input signals (LO and RF), which are mixed to
produce an IF output signal. In this case, HSPICE advanced analog analyses
offers two options:
■
Multi-tone HB analysis: specify the LO and RF base frequencies as two
separate tones on the .HB command.
■
Periodic AC analysis (HBAC): if one of the inputs is a small-signal, you can
use a faster linear analysis to analyze its effect. For example, if a mixer’s LO
is a large signal, but RF is a small signal, a single-tone HB analysis using
the LO frequency can be combined with HBAC in place of a 2-tone HB
analysis.
To demonstrate both techniques, this example analyzes an ideal mixer built
using behavioral elements. It is based on the demonstration netlist
mix_tran.sp, which is available in the directory: $installdir/demo/
hspice/rf_examples/.
* Ideal mixer example: transient analysis
.OPTIONS POST
vlo lo 0 1.0 sin (1.0 0.5 1.0g 0 0 90)
rrf1 rf1 rf 1.0
g1 0 if cur='1.0*v(lo)*v(rf)' $ mixer element
c1 0 if q='1.0e-9*v(lo)*v(rf)' $ mixer element
rout if ifg 1.0
vctrl ifg 0 0.0
h1 out 0 vctrl 1.0 $ convert I to V
rh1 out 0 1.0
vrf rf1 0 sin (0 0.001 0.8GHz 0 0 114)
.tran 10p 10n
.opt sim_accuracy=100
.end

This example uses behavioral-controlled current and charge sources to

simulate a mixer. The LO signal is driven by a 0.5 Volt sinusoid at 1 GHz and RF
is driven by a 10mV signal at 800 MHz. The mixer output is the voltage at node
out, v(out).

HSPICE® User Guide: Advanced Analog Simulation and Analysis 135

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

■
Two-tone HB Approach
■
HBAC Approach
■
Comparing Results

Two-tone HB Approach
To analyze this circuit using 2-tone HB, add:
■
HB source for LO: add HB 0.5 0 1 1 to the LO voltage source; this sets
the amplitude to 0.5, no phase shift for the first harmonic of the first tone,
which is 1 GHz.
■
HB source for RF: add HB 0.001 24 1 2 to the RF voltage source; this
sets the amplitude to 0.001, 24 degrees phase shift for the first harmonic of
the second tone (0.8 GHz).
■
An .HB command specifying both tones: .hb tones=1g 0.8g nharms=6
3; only a small number of harmonics are required to resolve the signals.
The complete mix_hb.sp netlist for 2-tone HB analysis is then:
* Ideal mixer example: 2-tone HB analysis
.OPTIONS POST
vlo lo 0 1.0 sin (1.0 0.5 1.0g 0 0 90) HB 0.5 0 1 1
rrf1 rf1 rf 1.0
g1 0 if cur='1.0*v(lo)*v(rf)' $ mixer element
c1 0 if q='1.0e-9*v(lo)*v(rf)' $ mixer element
rout if ifg 1.0
vctrl ifg 0 0.0
h1 out 0 vctrl 1.0 $ convert I to V
rh1 out 0 1.0
vrf rf1 0 sin (0 0.001 0.8GHz 0 0 114) HB 0.001 24 1 2
.opt sim_accuracy=100
.hb tones=1g 0.8g nharms=6 3
.end

This example is available in directory $<installdir>/demo/hspice/

rf_examples/.

HBAC Approach
To analyze this circuit using HBAC, start with the 2-tone HB analysis setup and
modify it as follows:

136 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

■
Replace the RF HB signal with an HBAC signal: change HB 0.001 24 1
2 to HBAC 0.001 24; this deactivates the source for HB and activates it for
HBAC with the same magnitude and phase.
■
Specify the frequency in the .HBAC command.
■
Change the .HB command to single tone:
.HB tones=1g nharms=6

HBAC takes care of the second tone.

■
Add a .HBAC command
.HBAC lin 1 0.8g 0.8g

This command runs an analysis at a single frequency point, 0.8 GHz. In

general, HBAC analysis can sweep the advanced analog frequency over a
range of values.
The following is the complete mix_hbac.sp netlist for HBAC analysis of this
simple mixer. This netlist also contains commands for performing periodic noise
analysis. It is available in the $installdir/demo/hspice/rf_examples/
directory.
* Ideal mixer example: HBAC analysis
.OPTIONS POST
vlo lo 0 1.0 sin (1.0 0.5 1.0g 0 0 90)
+ HB 0.5 0 1 1
rrf1 rf1 rf 1.0
g1 0 if cur='1.0*v(lo)*v(rf)' $ mixer element
c1 0 if q='1.0e-9*v(lo)*v(rf)' $ mixer element
rout if ifg 1.0
vctrl ifg 0 0.0
h1 out 0 vctrl 1.0 $ convert I to V
rh1 out 0 1.0
vrf rf1 0 sin (0 0.001 0.8GHz 0 0 114)
+ HBAC 0.001 24
.opt sim_accuracy=100
.hb tones=1g nharms=6
.hbac lin 1 0.8g 0.8g
* Noise analysis
.hbnoise v(out) rrf1 lin 40 0.1g 4g
.print hbnoise onoise nf
.probe hbnoise onoise nf
.end

HSPICE® User Guide: Advanced Analog Simulation and Analysis 137

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

Comparing Results
After running all three netlists above, you will have generated 3 output files:
■
mix_tran.tr0
■
mix_hb.hb0
■
mix_hbac.hb0
You can compare the results of the three analyses in Custom WaveView.
1. To run the netlists and start Custom WaveView, type:
hspice -i mix_tran.sp -o
hspice -i mix_hb.sp -o
hspice -i mix_hbac.sp -o
wv &

2. Use File > Import Waveform File and select the mix_tran.tr0,
mix_hb.hb0, and mix_hbac.hb0 files from the Open: Waveform Files
dialog box.
3. Select the v(out) signal from the mix_hb.hb0 file in the signal browser.
Double-click on the signal name or drag and drop the signal in the waveform.
You should see a frequency spectrum similar to the one from the
mix_hbac.hb0 file.
4. Select the v(out) signal from the mix_hbac.hb0 file in the signal browser.
Double-click on the signal name or drag and drop the signal in the waveform.
You should see a frequency spectrum similar to the one from the
mix_hb.hb0 file.

138 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

5. Convert the HB and HBAC histograms to time domain. In the waveform,

right-click in the name area of the panel containing the signal v(out), left-
click on the waveform label for v(out) from the mix_hb.hb0 file. From the
Panel menu, choose Signal 'v(out)' > To Time-Domain. To accept the
defaults for range and interval, click OK in the Convert to Time domain
window.
6. Repeat Step 5 for the v(out) signal from the mix_hbac.hb0 file.
7. Use Waveview > New to open a new waveform.
8. Select the v(out) signal from the mix_tran.tr0 file in the signal browser.
Double-click on the signal name or drag and drop the signal in the waveform.
9. Compare the time domain waveforms from the mix_hb.hb0 and
mix_hbac.hb0 files with the time domain waveform from the
mix_tran.tr0 file. In the file browser, click on IFT under derived
waveforms. The signals 0|v(out) and 1|v(out) should appear in the signal
browser. Select the 0|v(out) and 1|v(out) signals and drag and drop them in
the waveform. All three time domain signals should be displayed in the same
panel. The three signals are almost indistinguishable from each other.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 139

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

You can also use HBAC to perform noise analysis on advanced analog circuits
by using the .HBNOISE command, which is included in the mix_hbac.sp
netlist.
■
The .HBNOISE command invokes noise analysis, identifying an output
node where the noise is measured, an input noise source (in this case, rrf1)
which serves as a reference for noise figure computation and a frequency
sweep for the noise analysis.
■
The .PRINT and .PROBE hbnoise commands instruct HSPICE advanced
analog analyses to save the output noise and noise figure at each frequency
in the mix_hbac.printpn0 and mix_hbac.pn0 output files.
This ideal mixer is noiseless, except for the resistors at the input and output.
The mix_hbac.lis file contains detailed data on the individual noise source
contributions of the resistors. You can view mix_hbac.printpn0 to see the
output noise and noise figure at each frequency. In WaveView, you can view
mix_hbac.pn0 to plot the output noise and noise figure data as a function of
frequency.

140 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

Input Syntax
.HBAC frequency_sweep

Parameter Description

frequency_sweep Frequency sweep range for the input signal (also referred to as the input
frequency band or fin). You can specify LIN, DEC, OCT, POI, or
SWEEPBLOCK. Specify the nsteps, start, and stop frequencies using the
following syntax for each type of sweep:
■
LIN nsteps start stop
■
DEC nsteps start stop
■
OCT nsteps start stop
■
POI nsteps freq_values
■ SWEEPBLOCK=swblockname
■
DATA=dataname

HBAC Analysis Options

The following options directly relate to an HBAC analysis and override the
corresponding PAC options if specified in the netlist:
■
.OPTION HBACTOL, default = 1x10-8, Range = 1x10-14 to Infinity
■
.OPTION HBACKRYLOVDIM, default = 300, Range = 1 to Infinity
■ .OPTION HBACKRYLOVITER | HBAC_KRYLOV_ITER, default = 1000,
Range = 1 to Infinity
If these parameters are not specified, then the following conditions apply:
■ If HBACTOL > HBTOL, then HBACTOL = HBTOL
■
If HBACKRYLOVDIM < HBKRYLOVDIM, then HBACKRYLOVDIM =
HBKRYLOVDIM

Output Syntax
This section describes the syntax for the HBAC .PRINT and .PROBE
statements. These statements are similar to those used for HB analysis.
.PRINT and .PROBE Statements
.PRINT HB TYPE(NODES | ELEM)[INDICES]
.PROBE HB TYPE(NODES | ELEM)[INDICES]

HSPICE® User Guide: Advanced Analog Simulation and Analysis 141

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

Parameter Description

TYPE Specifies a harmonic type node or element.

TYPE can be one of the following:

■ Voltage type –
V = voltage magnitude and phase in degrees
VR = real component
VI = imaginary component
VM = magnitude
VP - Phase in degrees
VPD - Phase in degrees
VPR - Phase in radians
VDB - dB units
VDBM - dB relative to 1 mV
■
Current type –
I = current magnitude and phase in degrees
IR = real component
II = imaginary component
IM = magnitude
IP - Phase in degrees
IPD - Phase in degrees
IPR - Phase in radians
IDB - dB units
IDBM - dB relative to 1 mV
■
Power type – P
■
Frequency type – hertz[index], hertz[index1, index2, ...] You must specify the harmonic index for
the hertz variable. The frequency of the specified harmonics is dumped.

NODES | ELEM NODES or ELEM can be one of the following:

■ Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■
Current type – an element name (elemname)
■ Power type – a resistor (resistorname) or port (portname) element name
■
Frequency type – the harmonic index for the hertz variable. The frequency of the specified
harmonics is dumped.

INDICES Index to tones in the form [n1, n2, ..., nK, +/-1].
■ nj is the index of the j-th HB tone and the .HB statement contains K tones
■
+/-1 is the index of the HBAC tone

Wildcards are not supported if this parameter is used.

You can transform HB data into the time domain and output by using the following syntax:.PRINT
HBTRAN ov1 [ov2 ... ovN].PROBE HBTRAN ov1 [ov2 ... ovN]. See TYPE above for voltage and
current type definitions.

142 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance AC Analysis (.HBAC)

HBAC Output Data Files

An HBAC analysis produces these output data files:
■
Output from the .PRINT statement is written to a .printhb# file. This data
is against the input frequency band points.
• The header contains the large-signal fundamental frequency and the
range of small-signal frequencies.
• The columns of data are labeled as F(Hz), followed by the output
variable names. Each variable name has the associated mixing pair
value appended.
All N variable names and all M mixing pair values are printed for each
swept small-signal frequency value (a total of N*M for each frequency
value).
■
Output from the .PROBE statement is written to a .hb# file. This data is
against the input frequency band points.
■
Reported performance log statistics are written to a .lis file:
• Number of nodes
• Number of FFT points
• Number of equations
• Memory in use
• CPU time
• Maximum Krylov iterations
• Maximum Krylov dimension
• Target GMRES residual
• GMRES residual
• Actual Krylov iterations taken
• Frequency (swept input frequency values).

Using the .MEASURE Command with .HBAC

Since .HBAC requires an .HB analysis, the measure statements for this
analysis are the same as for .HB analysis. For example:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 143

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

.MEASURE HB result FIND out_var AT=val

Errors and Warnings

The following error and warning messages are used when HSPICE encounters
a problem with an HBAC analysis.

Error Messages
■
HBAC frequency sweep includes negative frequencies. HBAC allows only
frequencies that are greater than or equal to zero.
■
No HB statement is specified (error at parser). HBAC requires an HB
statement to generate the steady-state solution.

Warning Messages
■
More than one HBAC statement (warning at parser). HSPICE advanced
analog analyses uses only the last HBAC statement in the netlist.
■
No HBAC sources are specified (error at parser). HBAC requires at least
one HBAC source.
■
GMRES Convergence Failure. When GMRES (Generalized Minimum
Residual) reaches the maximum number of iterations and the residual is
greater than the specified tolerance. The HBAC analysis generates a
warning and then continues as if the data were valid. This warning reports
the following information:
• Final GMRES Residual
• Target GMRES Residual
• Maximum Krylov Iterations
• Actual Krylov Iterations taken

Shooting Newton AC Analysis (.SNAC)

You use the Shooting Newton AC (.SNAC) statement for analyzing linear
behavior in large-signal periodic systems. The .SNAC statement uses a
periodic AC (PAC) and Shooting Newton algorithm to perform linear analysis of

144 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

nonautonomous (driven) circuits, where the linear coefficients are modulated

by a periodic, steady-state signal.
The following section describes the periodic AC analysis based on a Shooting
Newton algorithm. This functionality is similar to the Harmonic Balance for
periodic AC (HBAC) analysis.
The following sections discuss these topics:
■
Prerequisites and Limitations
■
SNAC Example
■
Input Syntax
■
Output Syntax
■
SNAC Output Data Files
■
Using the .MEASURE Command with .SNAC
■
Errors and Warnings

Prerequisites and Limitations

The following prerequisites and limitations apply to SNAC:
■
Requires one and only one .SNAC statement. If you use multiple .SNAC
statements, HSPICE advanced analog analyses use only the last .SNAC
statement.
■
Requires one and only one .SN statement.
■
Requires placing the parameter sweep in the .SN statement.
■
Requires at least one periodic source.
■
Limited to simulations that can be reduced to a single tone SN analysis.
■
Supports unlimited number of sources.
■
The requested maximum harmonic in a .PROBE or .PRINT statement must
be less than or equal to half the number of harmonics specified in the SN
statement (that is, max_harm  nharms / 2).

HSPICE® User Guide: Advanced Analog Simulation and Analysis 145

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

SNAC Example
The following example is shipped with the HSPICE distribution as
mix_snac.sp and is available in the directory:
$installdir/demo/hspice/rf_examples/.
* Test SNAC: ideal I,Q mixer -rrd
.OPTIONS PROBE
.OPTIONS POST=2
$.OPTIONS snmaxiter=100
.OPTIONS SNACCURACY=5
vlo lo 0 1.0 cos(1.0 0.5 1g) $ Periodic, Large-Signal SN Input
rlo lo 0 1.0
rrf rf 0 1.0 $ Noise source
rrf1 rf1 rf 1.0 $ Noise source
g1 0 if cur='1.0*v(lo)*v(rf)' $ mixer element
c1 0 if q='1.0e-9*v(lo)*v(rf)' $ mixer element
rout if ifg 1.0
vctrl ifg 0 0.0
h1 out 0 vctrl 1.0
rh1 out 0 1.0
vrf rf1 0 snac .001 24.0 $ Small signal for SNAC with 1-tone SN Input
.sn tones=1.0g nharms=3
.snac lin 1 0.8g 0.8g
.print sn v(rf1) v(lo) v(out)
.print snfd v(rf1) v(lo) v(out)
.print snac v(rf1) v(lo) v(out)
.measure snac vout1 find v(out)[1,-1] at=0.8g
.measure snac vout2 find v(out)[0,1] at=0.8g
.measure snac vout3 find v(out)[1,1] at=0.8g
.measure sn vlo1 find v(lo) at=0.5n
.measure sn vlo2 find v(lo) at=1n
.measure snfd vlo3 find v(lo)[1] at=1
.end

146 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

Input Syntax
.SNAC frequency_sweep

Parameter Description

frequency_sweep Frequency sweep range for the input signal (also referred to as the input frequency band
or fin). You can specify LIN, DEC, OCT, POI, or SWEEPBLOCK. Specify the nsteps, start,
and stop frequencies using the following syntax for each type of sweep:
■ LIN nsteps start stop
■
DEC nsteps start stop
■ OCT nsteps start stop
■
POI nsteps freq_values
■
SWEEPBLOCK=swblockname
■
DATA=dataname

Output Syntax
This section describes the syntax for the SNAC .PRINT and .PROBE
statements. These statements are similar to those used for HB analysis.
.PRINT and .PROBE Statements
.PRINT SN TYPE(NODES | ELEM)[INDICES]
.PROBE SN TYPE(NODES | ELEM)[INDICES]

HSPICE® User Guide: Advanced Analog Simulation and Analysis 147

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

Parameter Description

TYPE Specifies a harmonic type node or element.

TYPE can be one of the following:

■ Voltage type –
V = voltage magnitude and phase in degrees
VR = real component
VI = imaginary component
VM = magnitude
VP - Phase in degrees
VPD - Phase in degrees
VPR - Phase in radians
VDB - dB units
VDBM - dB relative to 1 mV
■
Current type –
I = current magnitude and phase in degrees
IR = real component
II = imaginary component
IM = magnitude
IP - Phase in degrees
IPD - Phase in degrees
IPR - Phase in radians
IDB - dB units
IDBM - dB relative to 1 mV
■
Power type – P
■
Frequency type – hertz[index], hertz[index1, index2, ...] You must specify the harmonic index
for the hertz variable. The frequency of the specified harmonics is dumped.

NODES | NODES or ELEM can be one of the following:

ELEM ■ Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■
Current type – an element name (elemname)
■ Power type – a resistor (resistorname) or port (portname) element name
■
Frequency type – the harmonic index for the hertz variable. The frequency of the specified
harmonics is dumped.

INDICES Index to tones in the form [n1, +/-1].

■ n1 is the index of the SN tone
■
+/-1 is the index of the SNAC tone

Wildcards are not supported if this parameter is used.

You can transform SN data into the time domain and output by using the following syntax:.PRINT
SNTRAN ov1 [ov2 ... ovN]. PROBE SNTRAN ov1 [ov2 ... ovN]. See TYPE above for voltage and
current type definitions.

148 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton AC Analysis (.SNAC)

SNAC Output Data Files

A SNAC analysis produces these output data files:
■
Output from the .PRINT statement is written to a .printsnac# file.
■
This data is against the input frequency band points.
■
The header contains the large-signal fundamental frequency and the range
of small-signal frequencies.
■
The columns of data are labeled as F(Hz), followed by the output variable
names. Each variable name has the associated mixing pair value
appended. All N variable names and all M mixing pair values are printed for
each swept small-signal frequency value (a total of N*M for each frequency
value).
■
Output from the .PROBE statement is written to a .snac# file.
Reported performance log statistics are written to a .lis file:
■
Number of nodes
■
Number of FFT points
■
Number of equations
■
Memory in use
■
CPU time
■
Maximum Krylov iterations
■
Maximum Krylov dimension
■ Target GMRES residual
■
GMRES residual
■
Actual Krylov iterations taken
■
Frequency (swept input frequency values)

Using the .MEASURE Command with .SNAC

Since .SNAC requires an .SN analysis, the measure statements for this
analysis are the same as for .SN analysis. For example,
.MEASURE SN result FIND out_var AT=val

HSPICE® User Guide: Advanced Analog Simulation and Analysis 149

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Steady-State Voltage and Current Sources

Errors and Warnings

The following error and warning messages are used when HSPICE encounters
a problem with a SNAC analysis.

Error Messages
■
SNAC frequency sweep includes negative frequencies. SNAC allows only
frequencies that are greater than or equal to zero.
■
No SN statement is specified (error at parser). SNAC requires an SN
statement to generate the steady-state solution.

Warning Messages
■
More than one SNAC statement (warning at parser). HSPICE advanced
analog analyses uses only the last SNAC statement in the netlist.
■
No SNAC sources are specified (error at parser). SNAC requires at least
one SNAC source.
■
GMRES Convergence Failure. When GMRES (Generalized Minimum
Residual) reaches the maximum number of iterations and the residual is
greater than the specified tolerance. The SNAC analysis generates a
warning and then continue as if the data were valid. This warning reports the
following information:
• Final GMRES Residual
• Target GMRES Residual
• Maximum Krylov Iterations
• Actual Krylov Iterations taken

Steady-State Voltage and Current Sources

The I (current source) and V (voltage source) elements include extensions that
allow you to use them as sources of steady-state sinusoidal signals for HB/
HBAC and SN/SNAC analyses. When you use a power parameter to specify
the available power, you can also use these elements as power sources.
For a general description of the I and V elements, see Power Sources in the
HSPICE® User Guide: Elements.

150 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Steady-State Voltage and Current Sources

■
I and V Element Syntax
■
Examples

I and V Element Syntax

Vxxx p n
+ $ **** Voltage or Power Information ********
+ [[dc] mag] [ac [mag [phase]]] [HBAC [mag [phase]]]
+[SNAC [mag [phase]]]
+ [hb [mag [phase [harm [tone [modharm [modtone]]]]]]]
+ [transient waveform] [TRANFORHB=[1|0]]

+ $ **** Power Switch ********

+ [power=[0 | 1 | W | dbm]] [z0=val] [rdc=val] [rac=val]
+ [RHBAC=val] [rhb=val] [rtran=val]
Ixxx p n
+ $ **** Current or Power Information ********
+ [[dc] mag] [ac [mag [phase]]] [HBAC [mag [phase]]]
+ [SNAC [mag [phase]]]
+ [hb [mag [phase [harm [tone [modharm [modtone]]]]]]]
+ [transient waveform] [TRANFORHB=[1|0]]

+ $ **** Power Switch ********

+ [power=[0 | 1 | W | dbm]] [z0=val] [rdc=val] [rac=val]
+ [RHBAC=val] [rhb=val] [rtran=val]

Parameter Description

[[dc] mag] DC voltage or power source value. You do not need to specify DC
explicitly (default=0).

[ac [mag [phase]]] AC voltage or power source value.

[HBAC [mag [phase]]] Advanced analog HBAC voltage or power source value.

[SNAC [mag [phase]]] Advanced analog SNAC voltage or power source value.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 151

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Steady-State Voltage and Current Sources

Parameter Description

[hb [mag [phase [harm [tone Advanced analog HB voltage, current, or power source value. Multiple
[modharm [modtone]]]]]]] HB specifications with different harm, tone, modharm, and modtone
values are allowed.
■
phase is in degrees
■ harm and tone are indices corresponding to the tones specified in
the .HB statement. Indexing starts at 1 (corresponding to the first
harmonic of a tone).
■ modtone and modharm specify sources for multi-tone simulation. A
source specifies a tone and a harmonic, and up to 1 offset tone and
harmonic (modtone for tones and modharm for harmonics). The
signal is then described as:
V(or I) = mag*cos(2*pi*
(harm*tone+modharm*modtone)*t + phase)

[transient waveform] (Transient analysis) Any one of waveforms: AM, EXP, PULSE, PWL,
SFFM, or SIN. Multiple transient descriptions are not allowed.

[power=[0 | 1 | W | dbm]] (HSPICE advanced analog analyses) Power Switch

■ When 0 (default), element treated as a voltage or current source.
■
When 1 or W, element treated as a power source, realized as a
voltage source with a series impedance. In this case, the source
value is interpreted as RMS available power in units of Watts.
■
When dbm, element treated as a power source in series with the
port impedance. Values are in dbms.

You can use this parameter for transient analysis if the power source is
either DC or SIN.

[z0=val] (LIN analysis) System impedance used when converting to a power

source, inserted in series with the voltage source. Currently, this only
supports real impedance.
■
When power=0, z0 defaults to 0.
■ When power=1, z0 defaults to 50 ohms.

You can also enter zo=val.

[rdc=val] (DC analysis) Series resistance (overrides z0).

[rac=val] (AC analysis) Series resistance (overrides z0).

[RHBAC=val] (HSPICE advanced analog HBAC analysis) Series resistance

(overrides z0).

[rhb=val] (HSPICE advanced analog HB analysis) Series resistance (overrides

z0).

[rtran=val] (Transient analysis) Series resistance (overrides z0).

152 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Steady-State Voltage and Current Sources

Parameter Description

[TRANFORHB=[0|1]] ■ 0 (default): The transient description is ignored if an HB value is

given or a DC value is given. If no DC or HB value is given and
TRANFORHB=0, then HB treats the source as a DC source, and
the DC source value is the time=0 value.
■
1: HB analysis uses the transient description if its value is VMRF,
SIN, PULSE, PWL, or LFSR. If the type is a non-repeating PWL
source, then the time=infinity value is used as a DC source value.
For example, the following statement is treated as a DC source with
value=1 for HB:
v1 1 0 PWL (0 0 1n 1 1u 1) TRANFORHB=1
In contrast, the following statement is a 0V DC source:
v1 1 0 PWL (0 0 1n 1 1u 1) TRANFORHB=0
The following statement is treated as a periodic source with a 1us
period that uses PWL values:
v1 1 0 PWL (0 0 1n 1 0.999u 1 1u 0) R
TRANFORHB=1
To override the global TRANFORHB option, explicitly set
TRANFORHB for a V/I source.

Examples
Example 1
This example shows an HB source for a single tone analysis:
.hb tones=100MHz harms=7
I1 1 2 dc=1mA hb 3mA 0. 1 1

I1 is a current source with a the following time-domain description:

I1=1mA + 3mA*cos(2*pi*1.e8*t)

Example 2
This example shows HB sources used for a two-tone analysis:
.hb tones=1.e9 1.1e9 intmodmax=5
Vin lo 0 dc=0. hb 1.5 90 1 1
Vrf rf 0 dc=0. hb 0.2 0 1 2

These sources have the following time-domain descriptions:

Vin=1.5*cos(2*pi*1.e9*t - 90*pi/180) V
Vrf = 0.2*cos(2*pi*1.1e9*t) V

Example 3
The following HB source uses a modtone and modharms:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 153

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Phase Noise and Buffer Chains

.hb tones=2.e9 1.9e9 harms=5 5

Vm input gnd dc=0.5 hb 0.2 0. 1 1 -1 2

Vm has the following time-domain description:

Vm = 0.5 + cos(2*pi*1.e8*t)

Example 4
This example uses an HB source specified with a SIN source and TRANFORHB.
.hb tone=1.e8 harms=7
Vt 1 2 SIN(0.1 1.0 2.e8 0. 0. 90) tranforhb=1

Vt is converted to the following HB source:

Vt 1 2 dc=0.1 hb 1.0 0.0 2 1

Example 5
This example shows a power source (the units are Watts).
.hb tones=1.1e9 harms=9
Pt Input Gnd power=1 Z0=50. 1m 0. 1 1

Pt delivers 1 mW of power through a 50-ohm impedance.

Phase Noise and Buffer Chains

Phase noise is specific to oscillators. However, for buffer chains, you can do
periodic noise analysis. Phase noise that may be contributed by buffer,
amplifier, divider, or multiplier circuits is often referred to as the residual phase
noise. There are three commands in HSPICE advanced analog analyses that
can be used to help you predict such noise: .HBNOISE, .SNNOISE,
and .PTDNOISE. These analyses compute the noise spectral density at an
output variable taking into consideration the modulation effects.
.HBNOISE and .SNNOISE compute the average noise over one period,
assuming your circuit is driven by a periodic input signal.
.PTDNOISE computes the noise at one specific time point within the period, or
it can compute the noise as a function of time over one period. .PTDNOISE can
also convert the noise to a “jitter” value, which would be an uncertainty in the
timing as opposed to the uncertainty in the output voltage.

154 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

See the following sections for detailed information on these commands:

■
Multitone Harmonic Balance Noise (.HBNOISE)
■
Shooting Newton Noise Analysis (.SNNOISE)
■
Periodic Time-Dependent Noise Analysis (.PTDNOISE)

Multitone Harmonic Balance Noise (.HBNOISE)

An HBNOISE (Harmonic Balance noise) analysis simulates the noise behavior
in periodic systems and is designed for use with driven circuits. It employs a
PAC algorithm to perform noise analysis of non-autonomous (driven) circuits
under periodic, steady-state tone conditions. This can be extended to quasi-
periodic systems having more than one periodic, steady-state tone. One
application for a multitone HBNOISE analysis is determining mixer noise
figures under the influence of a strong interfering signal.
The PAC method simulates noise assuming that the stationary noise sources
and/or the transfer function from the noise source to a specific output are
periodically modulated.
■
The modulated noise source (thermal, shot, or flicker) is modeled as a
cyclostationary noise source.
■
A PAC algorithm solves for the modulated transfer function.
■
You can also use the HBNOISE PAC method with correlated noise sources
including the MOSFET level 9 and level 11 models and the behavioral noise
source in the G-element (voltage-dependent current source).
You use the .HBNOISE statement to perform a periodic noise analysis.
The following sections discuss these topics:
■
Supported Features
■
HBNOISE Example
■
Input Syntax
■
Output Syntax
■
Output Data Files
■
Measuring HBNOISE Analyses with .MEASURE

HSPICE® User Guide: Advanced Analog Simulation and Analysis 155

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

■
Optimizing with HBNOISE Measurement
■
Errors and Warnings

Supported Features
HBNOISE supports the following features:
■
All existing HSPICE advanced analog noise models.
■
Uses more than one single-tone, harmonic balance to generate the steady-
state solution.
■
Unlimited number of HB sources (using the same tone, possibly multiple
harmonics).
■
Includes stationary, cyclostationary, frequency-dependent, and correlated
noise effects.
■
Swept parameter analysis.
■
Results are independent of the number of HBAC sources in the netlist.

Prerequisites and Limitations

The following prerequisites and limitations apply to HBNOISE:
■
Requires one .HB statement (which determines the steady-state solution).
■
Requires at least one HB source or one TRANFORHB source.
■
Requires placing the parameter sweep in the .HB statement.
■
The requested maximum harmonic in .HBNOISE must be less than or equal
to half the number of harmonics used in harmonic balance (that is,
max_harm <= num_hb_harms/2).

HBNOISE Example
This example performs an HB analysis, then runs an HBNOISE analysis over a
range of frequencies, from 9.0e8 to 9.2e8 Hz. Simulation outputs the output
noise at V(out) and the single side-band noise figure versus the input frequency
band, from 1e8 to 1.2e8 Hz, to the *.pn0 file. The following is the netlist for this
example:

156 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

$$*-Ideal mixer + noise source

$ prints total noise at the output (1.57156p V/sqrt-Hz),
$ single-sideband noise figure, (3.01 dB)
$ double-sideband noise figure. (0 dB)
.OPTION PROBE
.OPTION POST=2
vlo lo 0 0.0 hb 1.0 0 1 1$ Periodic, HB Input
Ilo lo 0 0
rsrc rfin rf1 1.0 $ Noise source
c1 0 if q='1.0e-9*v(lo)*v(rfin)' $ mixer element
g1 0 if cur=’1.0*v(lo)*v(rfin)’ $ mixer element
rout if 0 1.0
vrf rf1 0 $ hbac 2.0 0.0
.hb tones=1.0g nharms=4 $ sweep mval 1 2 1
.HBNOISE rout rsrc lin 11 0.90g 0.92g
.print HBNOISE onoise ssnf dsnf
.probe HBNOISE onoise ssnf dsnf
.end

Input Syntax
.HBNOISE [output] [insrc] [parameter_sweep]
+ [n1, n2, ..., nk,+/-1]
+ [listfreq=(frequencies|none|all)] [listcount=val]
+ [listfloor=val] [listsources=on|off]

Parameter Description

output Output node, pair of nodes, or 2-terminal element. HSPICE advanced analog
analyses references equivalent noise output to this node (or pair of nodes).
Specify a pair of nodes as V(n+,n-). If you specify only one node, V(n+), then
HSPICE advanced analog analyses assumes that the second node is ground.
You can also specify a 2-terminal element name that refers to an existing element
in the netlist.

insrc An input source. If this is a resistor, HSPICE advanced analog analyses uses it as
a reference noise source to determine the noise figure. If the resistance value is
0, the result is an infinite noise figure.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 157

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

Parameter Description

parameter_sweep Frequency sweep range for the input signal. Also referred to as the input
frequency band or fin). You can specify LIN, DEC, OCT, POI, SWEEPBLOCK,
DATA, MONTE, or OPTIMIZE sweeps. Specify the nsteps, start, and stop
frequencies using the following syntax for each type of sweep:
■
LIN nsteps start stop
■ DEC nsteps start stop
■
OCT nsteps start stop
■ POI nsteps freq_values
■
SWEEPBLOCK=swblockname

n1,n2,...,nk, Index term defining the output frequency band (fout) at which the noise is
+/-1 evaluated. Generally,
fout=ABS(n1*f1+n2*f2+...+nk*fk+/-fin)
where:
■ f1,f2,...,fk are the first through k-th steady-state tones determined from the
harmonic balance solution
■
fin is the input frequency band defined by parameter_sweep.
The default index term is [1,1,...1,-1].
For a single tone analysis, the default mode (index term [1,-1]) is consistent with
simulating a high-side, down conversion mixer where the RF signal is specified by
the input frequency band and the noise is measured at a down-converted
frequency that the output frequency band specifies: fout=abs(fLO-fin). Where fLO
corresponds to the large signal specified by the first index 1 and fin is the input RF
signal specified by the second index -1.
In general, you can use the [n1,n2,...,nk,+/-1] index term to specify an arbitrary
offset. The noise figure measurement is also dependent on this index term.

listfreq Prints the element noise value to the .lis file. You can specify at which frequencies
the element noise value is printed. The frequencies must match the
sweep_frequency values defined in the parameter_sweep, otherwise they are
ignored.
In the element noise output, the elements that contribute the largest noise are
printed first. The frequency values can be specified with the NONE or ALL
keyword, which either prints no frequencies or every frequency defined in
parameter_sweep. Frequency values must be enclosed in parentheses. For
example:listfreq=(none)
listfreq=(all)
listfreq=(1.0G)
listfreq=(1.0G, 2.0G)The default value is NONE.

listcount Prints the element noise value to the .lis file, which is sorted from the largest to
smallest value. You do not need to print every noise element; instead, you can
define listcount to print the number of element noise frequencies. For
example, listcount=5 means that only the top 5 noise contributors are printed.
The default value is 1.

listfloor Prints the element noise value to the .lis file and defines a minimum meaningful
noise value (in V/Hz1/2 units). Only those elements with noise values larger than
listfloor are printed. The default value is 1.0e-14 V/Hz1/2.

158 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

Parameter Description

listsources Prints the element noise value to the .lis file when the element has multiple
noise sources, such as a FET, which contains the thermal, shot, and 1/f noise
sources. You can specify either ON or OFF: ON prints the contribution from each
noise source and OFF does not. The default value is OFF.

Output Syntax
The HSPICE advanced analog HB and SN noise analyses can output the
output noise (onoise), noise figures (NF, SSNF, and DSNF) and, the input
referred noise (inoise). This section describes the syntax for the HBNOISE
.PRINT and .PROBE statements.

.PRINT and .PROBE Statements

.PRINT HBNOISE [ONOISE] [NF] [SSNF] [DSNF] [INOISE]
.PROBE HBNOISE [ONOISE] [NF] [SSNF] [DSNF] [INOISE]

Parameter Description

ONOISE Outputs the voltage noise at the output frequency band across the output nodes in
the .HBNOISE statement. The data is plotted as a function of the input frequency band
points. Units are in V/Hz1/2. Simulation ignores ONOISE when applied to autonomous
circuits.

NF NF and SSNF both output a single-side band noise figure as a function of the input
SSNF frequency band points:

NF = SSNF = 10 Log(SSF)

Single side-band noise factor, SSF = {(Total Noise at output, at the output frequency
band, originating from all frequencies) - (Load Noise originating from the output
frequency band)} / (Input Source Noise originating from the input frequency band).

DSNF DSNF outputs a double side-band noise figure as a function of the input frequency band
points.

DSNF = 10 Log(DSF)

Double side-band noise factor, DSF = {(Total Noise at output, at the output frequency
band, originating from all frequencies) - (Load Noise originating from the output
frequency band)} / (Input Source Noise originating from the input frequency band and
from the image of the input frequency band).

INOISE Outputs input referred noise which can be printed, probed, or measured.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 159

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

Output Data Files

An HBNOISE analysis produces these output data files:
■
Output from the .PRINT statement is written to a .printpn# file.
■
Output from the .PROBE statement is written to a .pn# file.
Both the *.printpn# and *.pn# files output data against the input
frequency band points.
■
Standard output information is written to a .lis file:
• simulation time
• HBNOISE linear solver method
• HBNOISE simulation time
• total simulation time
See also Using Noise Analysis Results as Input Noise Sources.

Measuring HBNOISE Analyses with .MEASURE

Note: A .MEASURE HBNOISE statement cannot contain an expression
that uses a HBNOISE variable as an argument. Also, you cannot
use a .MEASURE HBNOISE statement for error measurement
and expression evaluation of HBNOISE.

The .MEASURE HBNOISE syntax supports several types of measurements:

■
Find-when
.MEASURE HBNOISE result FIND out_var1
+ AT = Input_Frequency_Band value

The previous measurement yields the result of a variable value at a specific

input frequency band point.
.MEASURE HBNOISE result FIND out_var1
+ WHEN out_var2 = out_var3

The previous measurement yields the result at the input frequency point
when out_var2 == out_var3.
.MEASURE HBNOISE result WHEN out_var2 = out_var3

160 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

The previous measurement yields the input frequency point when out_var2
== out_var3.
■ Average, RMS, min, max, and peak-to-peak
.MEASURE HBNOISE result [RMS] out_var [FROM = IFB1]
+ [TO = IFB2]
■
Integral evaluation
.MEASURE HBNOISE result INTEGRAL out_var
+ [FROM = IFB1] [TO = IFB2]

This measurement integrates the out_var value from the IFB1 frequency to
the IFB2 frequency.
■
Derivative evaluation
.MEASURE HBNOISE result DERIVATIVE out_var AT = IFB1

This measurement finds the derivative of out_var at the IFB1 frequency

point.
Note: .MEASURE HBNOISE cannot contain an expression that uses
an hbnoise variable as an argument. You also cannot
use .MEASURE HBNOISE for error measurement and
expression evaluation of HBNOISE.
■
Input referred noise
.MEASURE [HBNOISE|SNNOISE] result FIND inoise
+ AT = IFB_value

This measurement yields the result of the input referred noise at a specific
input frequency band point.
.MEASURE [HBNOISE|SNNOISE] result FIND inoise
+ WHEN out_var2 = out_var3

This measurement yields the result at the input frequency point when
out_var2 == out_var3.
.MEASURE HBNOISE result func inoise [FROM = IFB1]
+ [TO = IFB2]

Where func is one of the following measurement types:

• AVG (average): Calculates the area under the inoise curve, divided by
the periods of interest.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 161

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Noise (.HBNOISE)

• MAX (maximum): Reports the maximum value of inoise over the

specified interval.
• MIN (minimum): Reports the minimum value of inoise over the
specified interval.
• PP (peak-to-peak): Reports the maximum value minus the minimum
value of inoise over the specified interval.
• RMS (root mean squared): Calculates the square root of the area under
the inoise curve, divided by the period of interest.
.MEASURE HBNOISE result INTEGRAL inoise
+ [FROM =IFB1] [TO = IFB2]

This measurement integrates the inoise value from the IFB1 frequency to
the IFB2 frequency.
.MEASURE HBNOISE result DERIVATIVE inoise AT = IFB1

This measurement finds the derivative of inoise at the IFB1 frequency

point.
The HSPICE advanced analog analyses optimization flow can read the
measured data from a .MEASURE HBNOISE analysis. This flow can be
combined in the HSPICE advanced analog analyses optimization routine with a
.MEASURE HBTR analysis (see Using .MEASURE with .HB Analyses) and a
.MEASURE PHASENOISE analysis (see Measuring Phase Noise
with .MEASURE PHASENOISE).

Optimizing with HBNOISE Measurement

The required statements are:
■ Analysis statement
.HB TONES=f1[f2...fn] [NHARMS=h1 [,h2...hn]]
+ SWEEP OPTIMIZE=OPTxxx RESULT=measname MODEL=mname

For example:
.HBOSC tone=1g nharms=5 PROBENODE=out,gnd,0.8
+ SWEEP OPTIMIZE=opt1 RESULT=y1,y2 MODEL=m1
.MODEL m1 OPT level=0
■
Measure statement
.MEASURE HBNOISE measname FIND out_var1 AT=val GOAL=val

162 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Errors and Warnings

For .HBNOISE errors, see the list of .HBAC Errors and Warnings.

Shooting Newton Noise Analysis (.SNNOISE)

A Shooting Newton noise (SNNOISE) analysis simulates the noise behavior in
periodic systems. It uses a PAC algorithm to perform noise analysis of non-
autonomous (driven) circuits under periodic, steady-state tone conditions.
SNNOISE is similar to the HBNOISE analysis.
The PAC method simulates noise assuming that the stationary noise sources
and/or the transfer function from the noise source to a specific output are
periodically modulated.
■
The modulated noise source (thermal, shot, or flicker) is modeled as a
cyclostationary noise source.
■
A PAC algorithm solves the modulated transfer function.
■
You can also use the SNNOISE PAC method with correlated noise sources
including the MOSFET Level 9 and Level 11 models and the behavioral
noise source in the G-element (voltage-dependent current source).
You use the .SNNOISE statement to perform a periodic noise analysis.
The following sections discuss these topics:
■
Supported Features
■
SNNOISE Analysis Example
■ Input Syntax
■
Output Syntax
■ Output Data Files
■
Measuring SNNOISE Analyses with .MEASURE

Supported Features
SNNOISE supports the following features:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 163

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

■
All existing HSPICE advanced analog noise models.
■
Uses Shooting Newton to generate the steady-state solution.
■
Unlimited number of sources.
■
Includes stationary, cyclostationary, frequency-dependent, and correlated
noise effects.
■
Swept parameter analysis.
■
Results are independent of the number of SNAC sources in the netlist.

Prerequisites and Limitations

The following prerequisites and limitations apply to SNNOISE:
■
Requires one .SN statement (which determines the steady-state solution).
■
Requires at least one periodic source. Does not recognize HB sources.
■
Requires placing the parameter sweep in the .SN statement.

SNNOISE Analysis Example

This example performs an SN analysis, then runs an SNNOISE analysis over a
range of frequencies from 9.0e8 to 9.2e8 Hz. Simulation outputs the output
noise at V(out) and the single side-band noise figure versus the input frequency
band, from 9.0e8 to 9.2e8 Hz, to the *.pn0 file. The following is the netlist for
this example:

164 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

*
$$*-Ideal mixer + noise source
$ prints total noise PSD at the output (2.47e-20 V^2) when q=0
$ single-sideband noise figure, (3.01 dB)
$ double-sideband noise figure. (0 dB)
.OPTION PROBE
.OPTION POST=2
vlo lo 0 0.0 cos (0 1.0 1.0g 0 0 0)
Ilo lo 0 0
rsrc rfin rf1 1.0 $ Noise source
g1 0 if cur='1.0*v(lo)*v(rfin)' $ mixer element
c1 0 if q='1.0e-9*v(lo)*v(rfin)' $ mixer element
rout if 0 1.0
vrf rf1 0 $ hbac 2.0 0.0
.option delmax=0.002n
.SN tones=1G nharms=4 trstab=10n
.SNNOISE rout rsrc lin 11 0.90g 0.92g
.probe SNNOISE onoise ssnf dsnf
.print SNNOISE onoise ssnf dsnf
.end

Input Syntax
.SNNOISE [output] [insrc] [parameter_sweep]
+ [n1+/-1]
+ [listfreq=(frequencies|none|all)] [listcount=val]
+ [listfloor=val] [listsources=on|off]

Parameter Description

output Can be an output node, pair of nodes, or a 2-terminal element. HSPICE advanced analog
analysis references the equivalent noise output to this node (or pair of nodes). Specify a pair
of nodes as V(n+,n-). If you specify only one node V(n+), then HSPICE advanced analog
analysis assumes that the second node is ground. You can also specify a 2-terminal element
name that refers to an existing element in the netlist. If the 2-terminal element is a voltage
source, then HSPICE advanced analog analysis outputs the noise current through the voltage
source in A/sqrt(Hz). For all other 2-terminal devices, HSPICE outputs the noise voltage
across the device in V/sqrt(Hz).

insrc An input source. If this is a resistor, HSPICE advanced analog analyses uses it as a reference
noise source to determine the noise figure. If the resistance value is 0, the result is an infinite
noise figure.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 165

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Parameter Description

parameter_sweep Frequency sweep range for the input signal. Also referred to as the input frequency band or
fin). You can specify LIN, DEC, OCT, POI, SWEEPBLOCK, DATA, MONTE, or OPTIMIZE
sweeps. Specify the nsteps, start, and stop frequencies using the following syntax for each
type of sweep:
■
LIN nsteps start stop
■ DEC nsteps start stop
■
OCT nsteps start stop
■ POI nsteps freq_values
■
SWEEPBLOCK=swblockname

n1,+/-1 Index term defining the output frequency band at which the noise is evaluated. The output
frequency is computed according to fout=|n1*f1 +/- fin|, where f1 is the fundamental tone
(inverse of the fundamental period) and fin is from the frequency sweep.
For a single tone analysis, the default mode (index term [1,-1]) is consistent with simulating a
high-side, down conversion mixer where the RF signal is specified by the input frequency band
and the noise is measured at a down-converted frequency that the output frequency band
specifies: fout=abs(fLO-fin). Where fLO corresponds to the large signal specified by the first
index 1 and fin is the input RF signal specified by the second index -1.
See Specifying Variant Indices and Measuring SNNOISE Analyses with .MEASURE.

listfreq Prints the element noise value to the .lis file. You can specify at which frequencies the element
noise value is printed. The frequencies must match the sweep_frequency values defined in the
parameter_sweep, otherwise they are ignored.
In the element noise output, the elements that contribute the largest noise are printed first. The
frequency values can be specified with the NONE or ALL keyword, which either prints no
frequencies or every frequency defined in parameter_sweep. Frequency values must be
enclosed in parentheses. For example:
listfreq=(none)
listfreq=(all)
listfreq=(1.0G)
listfreq=(1.0G, 2.0G)

The default value is NONE.

listcount Prints the element noise value to the .lis file, which is sorted from the largest to smallest value.
You do not need to print every noise element; instead, you can define listcount to print the
number of element noise frequencies. For example, listcount=5 means that only the top 5
noise contributors are printed. The default value is 1.

listfloor Prints the element noise value to the .lis file and defines a minimum meaningful noise value
(in V/Hz1/2 units). Only those elements with noise values larger than listfloor are printed.
The default value is 1.0e-14 V/Hz1/2.

listsources Prints the element noise value to the .lis file when the element has multiple noise sources,
such as a FET, which contains the thermal, shot, and 1/f noise sources. You can specify either
ON or OFF: ON Prints the contribution from each noise source and OFF does not. The default
value is OFF.

166 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Specifying Variant Indices

.SNNOISE is the appropriate HSPICE advanced analog analysis for the
computation of noise at the output of a sample and hold circuit. When
using .SNNOISE, you need to specify the indices as [0,1] instead of the default
[1,-1]. When you specify the indices as [0,1], you get results that are “what you
see is what you get” with respect to the frequency sweep specified in
the .SNNOISE command. There are two more important things to consider
when using .SNNOISE: you must use enough harmonics to resolve the clock
edge and you will want a high density of SN time points. It is recommended that
the number of time points be between 2 and 20 times the number of harmonics
used. To increase the density of the time points during the SN analysis, you can
use the option DELMAX to specify a maximum time step.
For example, you can use the following settings:
.OPTION SNACCURACY=50
.OPTION DELMAX=5p

.SN TONES=5e6 nharms=2000 trinit=800n

.SNNOISE v(vo) vsin [0,1] dec 100 1e6 1e9

This example uses 100 points per decade for the frequency sweep, instead of
1000 points per decade. This is to speed up the simulation. It does not affect
the accuracy of the results.

Output Syntax
This section describes the syntax for the SNNOISE .PRINT and .PROBE
statements.

.PRINT and .PROBE Statements

.PRINT SNNOISE [ONOISE] [NF] [SSNF] [DSNF] [INOISE]
.PROBE SNNOISE [ONOISE] [NF] [SSNF] [DSNF] [INOISE]

Parameter Description

ONOISE Outputs the voltage noise at the output frequency band across the output nodes in
the .SNNOISE statement. The data is plotted as a function of the input frequency band
points. Units are in V/Hz1/2. Simulation ignores ONOISE when applied to autonomous
circuits.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 167

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Parameter Description

NF NF and SSNF both output a single-side band noise figure as a function of the input
SSNF frequency band points:
NF = SSNF = 10 Log(SSF)
Single side-band noise factor, SSF = {(Total Noise at output, at the output frequency
band, originating from all frequencies) - (Load Noise originating from the output
frequency band)} / (Input Source Noise originating from the input frequency band).

DSNF DSNF outputs a double side-band noise figure as a function of the input frequency band
points.
DSNF = 10 Log(DSF)
Double side-band noise factor, DSF = {(Total Noise at output, at the output frequency
band, originating from all frequencies) - (Load Noise originating from the output
frequency band)} / (Input Source Noise originating from the input frequency band and
from the image of input frequency band).

INOISE Outputs input referred noise which can be printed, probed, or measured.

Output Data Files

An SNNOISE analysis produces these output data files:
■
Output from the .PRINT statement is written to a .printsnpn# file.
■
Output from the .PROBE statement is written to a .snpn# file.
Both the *.printsnpn# and *.pn# files output data against the input
frequency band points.
■
Standard output information is written to a .lis file:
• simulation time
• SNNOISE linear solver method
• SNNOISE simulation time
• total simulation time
See also Using Noise Analysis Results as Input Noise Sources.

168 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Measuring SNNOISE Analyses with .MEASURE

Note: A .MEASURE SNNOISE statement cannot contain an expression
that uses a SNNOISE variable as an argument. Also, you cannot
use a .MEASURE SNNOISE statement for error measurement
and expression evaluation of SNNOISE.

The .MEASURE SNNOISE syntax supports four types of measurements:

■
Find-when
.MEASURE SNNOISE result FIND out_var1
+ At = Input_Frequency_Band value

The previous measurement yields the result of a variable value at a specific

input frequency band point.
.MEASURE SNNOISE result FIND out_var1
+ WHEN out_var2 = out_var3

The previous measurement yields the result at the input frequency point
when out_var2 == out_var3.
.MEASURE SNNOISE result WHEN out_var2 = out_var3

The previous measurement yields the input frequency point when out_var2
== out_var3.
■
Average, RMS, min, max, and peak-to-peak
.MEASURE SNNOISE result [RMS] out_var [FROM = IFB1]
+ [TO = IFB2]
■
Integral evaluation
.MEASURE SNNOISE result INTEGRAL out_var
+ [FROM = IFB1] [TO = IFB2]

This measurement integrates the out_var value from the IFB1 frequency to
the IFB2 frequency.
■
Derivative evaluation
.MEASURE SNNOISE result DERIVATIVE out_var AT = IFB1

This measurement finds the derivative of out_var at the IFB1 frequency

point.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 169

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Noise Analysis (.SNNOISE)

Note: .MEASURE SNNOISE cannot contain an expression that uses

an hbnoise variable as an argument. You also cannot
use .MEASURE SNNOISE for error measurement and
expression evaluation of SNNOISE.
■
Input referred noise:
.MEASURE SNNOISE result FIND inoise
+ AT = IFB_value

This measurement yields the result of the input referred noise at a specific
input frequency band point.
.MEASURE SNNOISE result FIND inoise
+ WHEN out_var2 = out_var3

This measurement yields the result at the input frequency point when
out_var2 == out_var3.
.MEASURE SNNOISE result func inoise [FROM = IFB1]
+ [TO = IFB2]

Where func is one of the following measurement types:

• AVG (average): Calculates the area under the inoise curve, divided by
the periods of interest.
• MAX (maximum): Reports the maximum value of inoise over the
specified interval.
• MIN (minimum): Reports the minimum value of inoise over the
specified interval.
• PP (peak-to-peak): Reports the maximum value minus the minimum
value of inoise over the specified interval.
• RMS (root mean squared): Calculates the square root of the area under
the inoise curve, divided by the period of interest.
.MEASURE SNNOISE result INTEGRAL inoise
+ [FROM =IFB1] [TO = IFB2]

This measurement integrates the inoise value from the IFB1 frequency to
the IFB2 frequency.
.MEASURE SNNOISE result DERIVATIVE inoise AT = IFB1

This measurement finds the derivative of inoise at the IFB1 frequency

point.

170 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Transfer Function Analysis (.HBXF)

Multitone Harmonic Balance Transfer Function

Analysis (.HBXF)
The .HBXF command calculates the transfer function from a given source in
the circuit to a designated output. Frequency conversion is calculated from the
input frequencies to a single output frequency that is specified with the
command. The relationship between the .HBXF command and the input/output
is expressed in the following equation:

Equation 35
Y m  j 0  =  HBXFm,n  j0 ,j   +     Xn  j   +   
W
where:
■
HBXF m ,n  j 0 ,j   +    is the transfer function from input port n to the
output port m
■
W is the set of all possible harmonics
■
 +  is the input frequency
■
 is the offset frequency
■
m is the output node number
■
n is the input node number
■
 0 is the output frequency
■
Y is the output (voltage or current)
■
X is the input (voltage or current)
The following sections discuss these topics:
■
Supported Features
■ Example
■
HBXF Test Listing
■
Input Syntax
■ Output Syntax
■
Output Data Files
■ Using the .MEASURE Command with .HBXF

HSPICE® User Guide: Advanced Analog Simulation and Analysis 171

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Transfer Function Analysis (.HBXF)

Supported Features
The .HBXF command supports the following features:
■
All existing HSPICE advanced analog models and elements
■
Sweep parameter analysis
■
Unlimited number of HB sources

Prerequisites and Limitations

The following prerequisites and limitations apply to the .HBXF command:
■
Only one .HBXF statement is required. If you use multiple .HBXF
statements, HSPICE advanced analog analyses use only the last .HBXF
statement.
■
At least one .HB statement is required, which determines the steady-state
solution.
■
Parameter sweeps must be placed in the .HB statements.

Example
Based on the HB analysis, the following example computes the trans-
impedance from isrc to v(1).
.hb tones=1e9 nharms=4
.hbxf v(1) lin 10 1e8 1.2e8
.print hbxf tfv(isrc) tfi(n3)

172 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Transfer Function Analysis (.HBXF)

HBXF Test Listing

* Test HBXF: nonlinear order-2 poly equation
.OPTIONS PROBE
.OPTIONS POST=2
vlo lo 0 cos(0 1.0 1g 0 0) tranforhb=1
rlo lo 0 50
vrf1 rf1 0 0
rrf1 rf1 0 50
E1 out 0 POLY(2) lo 0 rf1 0 0 1 1 1 10 1
rout out 0 50
.hb tones=1g nharms=5
.hbxf v(out) lin 2 100meg 200meg
.print hb v(out) v(rf1) v(lo)
.print hbxf tfv(vrf1) tfv(vlo)
.end

Input Syntax
.HBXF out_var freq_sweep

Parameter Description

out_var Specify i(2_port_elem) or V(n1[,n2])

freq_sweep Frequency sweep range for the input signal (also referred to as the input frequency
band (fin)). A sweep of type LIN, DEC, OCT, POI, or SWEEPBLOCK. Specify the
nsteps, start, and stop frequencies using the following syntax for each type of
sweep:
■ LIN nsteps start stop
■
DEC nsteps start stop
■
OCT nsteps start stop
■
POI nsteps freq_values
■
SWEEPBLOCK=swblockname
Specify the frequency sweep range for the output signal. HSPICE advanced
analog analyses determines the offset frequency in the input sidebands; for
example,
f1 = abs(fout - k*f0) s.t. f1<=f0/2
The f0 is the steady-state fundamental tone, and f1 is the input frequency.

Output Syntax
This section describes the syntax for the HBXF .PRINT and .PROBE
statements.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 173

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Multitone Harmonic Balance Transfer Function Analysis (.HBXF)

.PRINT and .PROBE Statements

.PRINT HBXF TYPE (NODES | ELEM)
.PROBE HBXF TYPE (NODES | ELEM)

Parameter Description

TYPE TYPE can be one of the following:

■ TFV = existing source
■
TFI = placeholder value for the current source attached to the given node.
The transfer function is computed on the output variables and input current or
voltage.

NODES | ELEM NODES or ELEM can be one of the following:

■
Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■ Current type – an element name (elemname)
■
Power type – a resistor (resistorname) or port (portname) element name.

Output Data Files

An HBXF calculation produces these output data files:
■
Output from the .PRINT statement is written to a .printxf# file.
• The output is in ohms, siemens, or undesignated units, and the header
in the output file is Z(..). Y(..) or GAIN(..).
■
Output from the .PROBE statement is written to an
.xf# file.
■
Reported performance log statistics are written to a .lis file:
• HBXF CPU time
• HBXF peak memory usage

Using the .MEASURE Command with .HBXF

Since .HBXF requires an .HB analysis, the measure statements for this
analysis are the same as for .HB analysis. For example,
.MEASURE HB result FIND out_var AT=val

174 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Transfer Function Analysis (.SNXF)

Shooting Newton Transfer Function Analysis (.SNXF)

The .SNXF command calculates transfer functions from an arbitrary number of
small signal sources to a designated output in a circuit under periodic steady-
state conditions. Frequency conversion is calculated from multiple input
frequencies to a single output at a single frequency that is specified on the
command line.
The following sections discuss these topics:
■ Prerequisites and Limitations
■
Example
■
SNXF Test Listing
■
Input Syntax
■
Output Syntax
■
Output Data Files
■
Using the .MEASURE Command with .SNXF

Prerequisites and Limitations

The following prerequisites and limitations apply to the .SNXF command:
■
Only one .SNXF statement is required. If you use multiple .SNXF
statements, HSPICE advanced analog analyses uses only the last one.
■
At least one .SN statement is required, which determines the steady-state
solution.
■
Parameter sweeps must be placed in the .SN statements.

Example
Based on the SN analysis, the following example computes the
transimpedance from isrc to v(1).
.SN tones=1e9 nharms=4
.SNXF v(1) lin 10 1e8 1.2e8
print SNXF TFV(isrc) TFI(n3)

HSPICE® User Guide: Advanced Analog Simulation and Analysis 175

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Transfer Function Analysis (.SNXF)

SNXF Test Listing

* Test SNXF: nonlinear order-2 poly equation

.OPTIONS PROBE
.OPTIONS POST=2
vlo lo 0 cos(0 1.0 1g 0 0)
rlo lo 0 50
vrf1 rf1 0 0
rrf1 rf1 0 50
E1 out 0 POLY(2) lo 0 rf1 0 0 1 1 1 10 1
rout out 0 50
.opt delmax=.01n
.sn tones=1g nharms=5
.snxf v(out) lin 2 100meg 200meg
.print sn v(out) v(rf1) v(lo)
.print snxf tfv(vrf1) tfv(vlo)
.end

Input Syntax
.SNXF out_var freq_sweep

176 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Transfer Function Analysis (.SNXF)

Parameter Description

Parameter Description

out_var Specify i(2_port_elem) or V(n1[,n2])

freq_sweep Frequency sweep range for the input signal (also referred to as the input frequency
band (fin)). A sweep of type LIN, DEC, OCT, POI, or SWEEPBLOCK. Specify the
nsteps, start, and stop frequencies using the following syntax for each type of
sweep:
■
LIN nsteps start stop
■
DEC nsteps start stop
■
OCT nsteps start stop
■
POI nsteps freq_values
■ SWEEPBLOCK=swblockname
Specify the frequency sweep range for the output signal. HSPICE advanced
analog analyses determines the offset frequency in the input sidebands Fin, where
Fin = abs(n*F0 +/- Fout). F0 is the steady-state fundamental tone, and Fout is the
output frequency. SNXF then generates the transfer functions from all of the input
sidebands (the Fin values) to the output frequency Fout.

Output Syntax
This section describes the syntax for the SNXF .PRINT and .PROBE
statements.
.PRINT and .PROBE Statements
.PRINT SNXF TYPE(NODES | ELEM)
.PROBE SNXF TYPE(NODES | ELEM)
Parameter Description
TYPE can be one of the following:
■ TFV = existing source
■
TFI = placeholder value for the current source attached to the given node.
The transfer function is computed on the output variables and input current or
voltage .NODES | ELEM NODES or ELEM can be one of the following:
■
Voltage type – a single node name (n1), or a pair of node names, (n1,n2)
■
Current type – an element name (elemname)
■ Power type – a resistor (resistorname) or port (portname) element name

HSPICE® User Guide: Advanced Analog Simulation and Analysis 177

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
Shooting Newton Transfer Function Analysis (.SNXF)

Output Data Files

An SNXF calculation produces these output data files:
■
Output from the .PRINT statement is written to a .printsnxf# file. The
output is in ohms, siemens, or undesignated units, and the header in the
output file is Z(..). Y(..) or GAIN(..).
■
Output from the .PROBE statement is written to an .snxf# file.
Reported performance log statistics are written to a .lis file:
■
SNXF CPU time
■
SNXF peak memory usage

Using the .MEASURE Command with .SNXF

Since .SNXF requires an .SN analysis, the measure statements for this
analysis are the same as for .SN analysis. For example,
.MEASURE SN result FIND out_var AT=val

178 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
References

References
[1] S. Maas, Nonlinear Microwave Circuits, Chapter 3, IEEE Press, 1997.
[2] R. Gilmore and M.B. Steer, “Nonlinear Circuit Analysis Using the Method of
Harmonic Balance - A Review of the Art, Part I, Introductory Concepts.”
International Journal of Microwave and Millimeter-wave Computer-Aided
Engineering, Volume 1, No. 1, pages 22-37, 1991.
[3] R. Gilmore and M.B. Steer, “Nonlinear Circuit Analysis Using the Method of
Harmonic Balance - A Review of the Art. Part II. Advanced Concepts.”
International Journal of Microwave and Millimeter-wave Computer-Aided
Engineering, Volume 1, No. 2, pages 159-180, 1991.
[4] V. Rizzoli, F. Mastri, F. Sgallari, G. Spaletta, “Harmonic-Balance Simulation
of Strongly Nonlinear Very Large-Size Microwave Circuits by Inexact
Newton Methods,” MTT-S Digest, pages 1357-1360, 1996.
[5] S. Skaggs, Efficient Harmonic Balance Modeling of Large Microwave
Circuits, Ph.D. thesis, North Carolina State University, 1999.
[6] R.S. Carson, High-Frequency Amplifiers, 2nd Edition, John Wiley & Sons,
1982
[7] S.Y. Liao, Microwave Circuit Analysis and Amplifier Design, Prentice-Hall,
1987.
[8] Y. Saad, Iterative Methods for Sparse Linear Systems, PWS Publishing
Company, 1995.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 179

N-2017.12
Chapter 7: Large Signal AC, Transfer Function, and Noise Analyses
References

180 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 8
8 Large Signal S-parameter Analyses

Describes how to do frequency translation and large-signal S-parameter

extraction, as well as noise parameter calculation.

This chapter discusses various techniques supported in HSPICE advanced

analog analyses for extracting circuit scattering parameters. Since advanced
analog circuits can operate under large-signal and small-signal conditions,
there are several types of scattering parameters that are useful to measure.
Linear small-signal scattering parameters represent the RF frequency-domain
transfer characteristics for a circuit that is operating at its DC bias condition, but
the stimulus and response signals are sufficiently small that they do not
influence the operating point. This type of analysis is performed using the .LIN
analysis. For information on doing small-signal S-parameter analysis (.LIN),
see Linear Network Parameter Analysis in the HSPICE User Guide: Signal
Integrity Modeling and Analysis.
In the case of RF mixers and receiver front-ends, some of the input and output
frequencies of interest involve a frequency translation. This translation is
intentional and caused by nonlinear mixing in the circuit due to devices being
driven by large-signal periodic waveforms. This type of scattering parameter
analysis therefore must begin by solving the large-signal periodic response and
then finding the small-signal behavior about this large-signal operating point.
This capability is provided by the .HBLIN analysis which has setup and
analysis control options similar to .LIN, but is capable of extracting
S-parameters about a large-signal periodic steady-state operating point.
In the case of circuits such as power amplifiers, the extraction of scattering
parameters is also important, but the circuit stimulus and response signals may
themselves be large-signal periodic waveforms. And, it can be important to
analyze how these S-parameters vary as a function of input power levels. This
capability is provided by the .HBLSP large-signal S-parameter analysis, which
uses large-signal stimulus for the S-parameter extractions.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 181

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

The following topics are presented in this chapter:

■
Frequency Translation S-Parameter (HBLIN) Extraction
■
Large-Signal S-parameter (HBLSP) Analysis

Frequency Translation S-Parameter (HBLIN) Extraction

Frequency translation scattering parameter (S-parameter) extraction is used to
describe N-port circuits that exhibit frequency translation effects, such as
mixers. The analysis is similar to the existing LIN analysis, except that the
circuit is first linearized about a periodically-varying operating point instead of a
simple DC operating point. After the linearization, the S-parameters between
circuit ports that convert signals from one frequency band to another are
calculated.
You use the .HBLIN statement to extract frequency translation S-parameters
and noise figures.
Frequency translation S-parameter describes the capability of periodically
linear time varying systems to shift signals in frequency. The S-parameters for
a frequency translation system are similar to the S-parameters of a linear-time-
varying system. They are defined as:

Equation 36
b = Sa
b i,m   
S i,j;m,n    = -------------------
a j,n   
a k  j p  n    = 0
The incident waves, a i ,n    and the reflected waves, b i ,n    , are defined using
these equations:

182 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

Equation 37
V i   + n 0  + Z 0i I i   + n 0 
a i ,n    = ---------------------------------------------------------------------------
2 Z 0i
V i   + n 0  – Z 0i I i   + n 0 
b i ,n    = ---------------------------------------------------------------------------
2 Z 0i
where:
■
 0 is the fundamental frequency (tone).
■
n is a signed integer.
■
i is the port number.
■ a i ,n    is the input wave at the frequency  + n 0 on the ith port.
■
b i ,n    is the reflected wave at the frequency  + n 0 on the ith port.
■
V i   + n 0  is the Fourier coefficient at the frequency  + n 0 of the
voltage at port i.
■
I i   + n 0  is the Fourier coefficient at the frequency  + n 0 of the current
at port i.
■ Z 0i is the reference impedance at port i.
■
V and I definitions are Fourier coefficients rather than phasors.
For a multi-tone analysis, they can be expressed as:

Equation 38
b = Sa
b i ,m ,m ...m   
Si    = --------------------------------------
1 2 N
-
,j;m 1 ...m N ,n 1 ,n 2 ...n N a j ,n ,n ...n   
1 2 N

a k ,p ,p ...p k  j ,p  n    = 0
1 2 N q q

HSPICE® User Guide: Advanced Analog Simulation and Analysis 183

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

Equation 39
N N
   
   n j  j
Vi  +
 
n j  j + Z 0i I i  +
   
 j=1   j=1 
a i ,n ,n ...n    = --------------------------------------------------------------------------------------------------
-
1 2 N
2 Z oi
N N
   
   n j  j
Vi  +
 
n j  j – Z 0i I i  +
   
 j=1   j=1 
b i ,n ,n ...n    = ------------------------------------------------------------------------------------------------
1 2 N
2 Z oi
where:
■
 j is the ith tone.
The frequency translation S-parameters are calculated by applying different
n j  j = 1  N  to different ports.
Limitations
The HBLIN analysis has these known limitations:
■
Noise parameters are not calculated for a mixed-mode operation.
■
Only the S-parameters corresponding to the set of frequencies specified at
each port are extracted.
■
Multiple small-signal tones are not supported.
■
The port element impedance cannot be specified as a complex number.
The following sections discuss these topics:
■ HB Analysis
■
Port Element
■ HBLIN Analysis
■ Output Syntax
■
Output Data Files

184 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

HB Analysis
An HB analysis is required prior to an HBLIN analysis. To extract the frequency
translation S-parameters, a sweep of the small-signal tone is necessary. You
can identify the small-signal tone sweep in the .HBLIN command or in the .HB
command together with a SS_TONE specification.
For additional information regarding HB analysis, see Harmonic Balance
Analysis on page 28.

Port Element
You must use a port (P) element as the termination at each port of the system.
To indicate the frequency band that the S-parameters are extracted from, it is
necessary to specify a harmonic index for each P-element.

Port Element Syntax

Without SS_TONE
Pxxx p n n_ref PORT=portnumber
+ [HBLIN = [H1, H2, ... HN, +/-1]] ...
With SS_TONE
Pxxx p n n_ref [PORT=portnumber]
+ [HBLIN = [H1, H2, ... +/-1 ... HN]] ...

Parameter Description

n_ref Reference node used when a mixed-mode port is specified.

PORT The port number. Numbered sequentially beginning with 1 with no shared port
numbers.

HBLIN Integer vector that specifies the harmonic index corresponding to the tones defined in
the .HB command. The +/-1 term corresponds to the small-signal tone specified by
SS_TONE in the .HB command. If there is no SS_TONE in the .HB command, the +/-
1 term must be at the last entry of the HBLIN vector.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 185

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

HBLIN Analysis
You use the .HBLIN statement to extract frequency translation S-parameters
and noise figures.
■
Input Syntax
■
Examples

Input Syntax
Without SS_TONE
.HBLIN frequency_sweep
+ [NOISECALC = [1|0|yes|no]] [FILENAME=file_name]
+ [DATAFORMAT = [ri|ma|db]]
+ [MIXEDMODE2PORT = [dd|cc|cd|dc|sd|sc|cs|ds]]
With SS_TONE
.HBLIN [NOISECALC = [1|0|yes|no]] [FILENAME=file_name]
+ [DATAFORMAT = [ri|ma|db]]
+ [MIXEDMODE2PORT = [dd|cc|cd|dc|sd|sc|cs|ds]]

Parameter Description

frequency_sweep Frequency sweep range for the input signal (also referred to as the input
frequency band or fin). You can specify LIN, DEC, OCT, POI, or
SWEEPBLOCK. Specify the nsteps, start, and stop frequencies using the
following syntax for each type of sweep:
■ LIN nsteps start stop
■
DEC nsteps start stop
■
OCT nsteps start stop
■
POI nsteps freq_values
■
SWEEPBLOCK=swblockname
■ DATA=dataname

NOISECALC Enables calculating the noise figure. The default is no (0).

FILENAME Specifies the output file name for the extracted S-parameters or the object
name after the -o command-line option. The default is the netlist file name.

DATAFORMAT Specifies the format of the output data file.

■ dataformat=RI, real-imaginary.
■
dataformat=MA, magnitude-phase. This is the default format for
Touchstone files.
■
dataformat=DB, DB(magnitude)-phase.

186 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

Parameter Description

MIXEDMODE2PORT Describes the mixed-mode data map of output mixed mode S-parameter
matrix. The availability and default value for this keyword depends on the first
two port (P element) configuration as follows:
■
case 1: p1=p2=single-ended (standard-mode P element)
available: ss
default: ss
■
case 2: p1=p2=balanced (mixed-mode P element)
available: dd, cd, dc, cc
default: dd
■
case 3: p1=balanced p2=single-ended
available: ds, cs
default: ds
■
case 4: p1=single p2=balanced
available: sd, sc
default: sd

Examples
Example 1
Single-tone analysis with frequency translation example. In this example, the 2-
port S-parameters from RF (1G-del_f) to IF (del_f) are extracted. The LO signal
is specified by normal voltage source Vlo. The frequency on port 1 is in the RF
band, 1G-del_f, and the frequency on port 2 is in the IF band, del_f. The IF
band is swept from 0- to 100-MHz. The results are output to the file ex1.s2p.
p1 RFin gnd port=1 HBLIN=(1,-1)
p2 IFout gnd port=2 HBLIN=(0,1)
Vlo LOin gnd DC 0 HB 2.5 0 1 1
.HB tones=1G harms=5
.HBLIN lin 5 0 100meg noisecalc=no filename=ex1
+ dataformat=ma

Example 2
Another single-tone analysis with frequency translation example. In this
example, the 3-port S-parameters are extracted. Port 3 provides the periodic
large signal. The frequency on port 1 is del_f, the frequency on port 2 is
1G*2-del_f, and the frequency on port 3 is 1G*1+del_f. The small-signal
frequency is swept from 0 to 100MHz. HBNOISE calculation is required. The
results are output to the file ex2.s3p.
p1 1 0 port=1 HBLIN=(0, 1)
p2 2 0 port=2 HBLIN=(2, -1)
p3 3 0 port=3 hb 0.5 0 1 1 HBLIN=(1, 1)
.HB tones=1G harms=5
.HBLIN lin 5 0 100meg noisecalc=yes filename=ex2

HSPICE® User Guide: Advanced Analog Simulation and Analysis 187

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Frequency Translation S-Parameter (HBLIN) Extraction

Output Syntax
This section describes the syntax for the HBLIN .PRINT and .PROBE
statements.

.PRINT and .PROBE Statements

.PRINT HBLIN Smn | Smn(TYPE) | S(m,n) | S(m,n)(TYPE)
.PROBE HBLIN Smn | Smn(TYPE) | S(m, n) | S(m, n)(TYPE)
.PRINT HBLIN SXYmn | SXYmn(TYPE) | SXY(m,n) | SXY(m,n)(TYPE)
.PROBE HBLIN SXYmn | SXYmn(TYPE) | SXY(m, n) | SXY(m, n)(TYPE)
.PRINT HBLIN NF SSNF DSNF
.PROBE HBLIN NF SSNF SSNF

Parameter Description

Smn | Smn(TYPE) | Complex 2-port parameters. Where:

S(m,n) | S(m,n)(TYPE) ■
m = 1 or 2
SXYmn | SXYmn(TYPE) | ■
n = 1 or 2
SXY(m,n) | SXY(m,n)(TYPE) ■ X and Y are used for mixed-mode S-parameter output.
■
The values for X and Y can be D (differential), C (common), or
S (single-end).
TYPE = R, I, M, P, PD, D, DB, or DBM
■ R = real
■
I = imaginary
■ M = magnitude
■
P = PD = phase in degrees
■
D = DB = decibels
■
DBM = decibels per 1.0e-3

NF NF and SSNF both output a single side-band noise figure as a

SSNF function of the input frequency band points:
NF = SSNF = 10 Log(SSF)
Single side-band noise factor, SSF = {(Total Noise at output, at the
output frequency band, originating from all frequencies) - (Load
Noise originating from the output frequency band)} / (Input Source
Noise originating from input frequency band).

DSNF DSNF outputs a double side-band noise figure as a function of the

input frequency band points.
DSNF = 10 Log(DSF)
Double side-band noise factor, DSF = {(Total Noise at output, at the
output frequency band, originating from all frequencies) - (Load
Noise originating from the output frequency band)} / (Input Source
Noise originating from the input frequency band and from the image
of input frequency band).

188 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

Output Data Files

An HBLIN analysis produces these output data files:
■
The S-parameters from the .PRINT statement are written to a .printhl#
file.
■
The extracted S-parameters from the .PROBE statement are written to a
.hl# file.

Large-Signal S-parameter (HBLSP) Analysis

Use the .HPSLP command to invoke periodically-driven nonlinear circuit
analyses for power-dependent S-parameters. An HBLSP analysis provides
three kinds of analyses for periodically-driven nonlinear circuits, such as those
that employ power amplifiers and filters:
■
Two-port power-dependant (large-signal) S-parameter extraction
■
Two-port small-signal S-parameter extraction
■
Two-port small-signal noise parameter calculation
Unlike small-signal S-parameters, which are based on linear analysis, power-
dependent S-parameters are based on the harmonic balance simulation. Its
solution accounts for nonlinear effects such as compression and variation in the
power levels.
The definition for power-dependent S-parameters is similar to that for small-
signal parameters. Power-dependent S-parameters are defined as the ratio of
reflected and incident waves by using this equation:
b=S*a ; S[i, j]=b[i,n]/a[j,n] when a[k,n](k!=j)=0
The incident waves, a[i, n], and reflected waves, b[i, n], are defined by using
these equations:
a[i, n] = (V[i](n*W0) + Zo[i] * I[i](n*W0)) / (2 * sqrt(Zo[i]))
b[i, n] = (V[i](n*W0) - Zo[i] * I[i](n*W0)) / (2 * sqrt(Zo[i]))
Where:
■ W0 is the fundamental frequency (tone).
■
n is a signed integer.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 189

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

■
i is the port number.
■ a[i, n] is the input wave at the frequency n*W0 on the ith port.
■
b[i, n] is the reflected wave at the frequency n*W0 on the ith port.
■ V[i](n*W0) is the Fourier coefficient at the frequency n*W0 of the voltage at
port i.
■
I[i](n*W0) is the Fourier coefficient at the frequency n*W0 of the current at
port i.
■ Zo[i] is the reference impedance at port i.
An HBLSP analysis only extracts the S-parameters on the first harmonic (that
is, n=1).
■
Input Syntax
■ Examples
■
Output Syntax
■
Output Data Files
■ HBLSP Analysis Limitations

Input Syntax
.HBLSP NHARMS=nh [POWERUNIT=[dbm|watt]]
+ [SSPCALC=[1|0|YES|NO]] [NOISECALC=[1|0|YES|NO]]
+ [FILENAME=file_name] [DATAFORMAT=[ri|ma|db]]
+ FREQSWEEP freq_sweep POWERSWEEP power_sweep

Parameter Description

NHARMS Number of harmonics in the HB analysis triggered by the .HBLSP statement.

POWERUNIT Power unit. Default is watt.

SSPCALC Extract small-signal S-parameters. Default is 0 (NO).

NOISECALC Perform small-signal 2-port noise analysis. Default is 0 (NO).

FILENAME Output data .p2d filename. Default is the netlist name or the object name after the
-o command-line option.

190 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

Parameter Description

DATAFORMAT Format of the output data file. Default is ma (magnitude, angle).

FREQSWEEP Frequency sweep specification. A sweep of type LIN, DEC, OCT, POI, or
SWEEPBLOCK. Specify the nsteps, start, and stop times using the following
syntax for each type of sweep:
■ LIN nsteps start stop
■
DEC nsteps start stop
■ OCT nsteps start stop
■
POI nsteps freq_values
■
SWEEPBLOCK=swblockname

This keyword must appear before the POWERSWEEP keyword.

POWERSWEEP Power sweep specification. A sweep of type LIN, DEC, OCT,POI, or

SWEEPBLOCK. Specify the nsteps, start, and stop frequencies using the
following syntax for each type of sweep:
■
LIN nsteps start stop
■
DEC nsteps start stop
■ OCT nsteps start stop
■
POI nsteps power_values
■ SWEEPBLOCK=swblockname

This keyword must follow the FREQSWEEP keyword.

Note: The FREQSWEEP and POWERSWEEP keywords must appear at the

end of an .HBLSP statement.

Examples
The following example does a 2-port single-tone, power-dependent S-
parameter extraction, without frequency translation:
■ Frequency sweep: The fundamental tone is swept from 0 to 1G
■
Power sweep: The power input at port 1 is swept from 6 to 10 Watts.
■ Five harmonics are required for the HB analysis. Large-signal S-parameters
are extracted on the first harmonic.
■
Five harmonics are required in the HBLSP triggered HB analysis.
■ The DC value in p1 statement is used to set the DC bias, which is used to
perform small-signal analyses.
■ Small-signal S-parameters are required extracted.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 191

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

■
Small-signal two-port noise analysis is required.
■
The data will be output to the ex1.p2d file.

Example 19 2-Port, Single Tone

p1 1 0 port=1 dc=1v
p2 2 0 port=2
.hblsp nharms=5 powerunit = watt
+ sspcalc=1 noisecalc=1 filename=ex1
+ freqsweep lin 5 0 1G powersweep lin 5 6 10

The following example generates large scale S-parameters as a function of

input for a differential equalizer.

Example 20 4-Port Network

* hblsp example
.opt post
p1 n1 0 port=1 ac=1
p2 n2 0 port=2
*** put your DUT
R1 n1 n2 10***
.hblsp nharms=5
+ freqsweep lin 4 1k 10k
+ powersweep lin 2 5 10
.end

Output Syntax
This section describes the syntax for the HBLSP .PRINT and .PROBE
statements. These statements only support S and noise parameter outputs.
Node voltage, branch current, and all other parameters are not supported in
HBLSP .PRINT and .PROBE statements.

.PRINT and .PROBE Statements

.PRINT HBLSP Smn | Smn(TYPE) | S(m, n) | S(m, n)(TYPE)
+ ...small signal 2-port noise params...
.PROBE HBLSP Smn | Smn(TYPE) | S(m, n) | S(m, n)(TYPE)

192 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

+ ...small signal 2-port noise params...

Parameter Description

Smn | Smn(TYPE) | Complex 2-port parameters. Where:

S(m,n) | S(m,n)(TYPE) ■
m = 1 or 2
■
n = 1 or 2
■
TYPE = R, I, M, P, PD, D, DB, or DBM
R = real
I = imaginary
M = magnitude
P = PD = phase in degrees
D = DB = decibels
DBM = decibels per 1.0e-3

... small signal 2-port noise G_AS | NF | RN | YOPT | GAMMA_OPT | NFMIN |

parameters ... VN2 | ZCOR | GN | RHON | YCOR | ZOPT | IN2

For a description of these parameters, see Linear Network

Parameter Analysis in the HSPICE User Guide: Signal Integrity
Modeling and Analysis.

Output Data Files

An HBLSP analysis produces these output data files:
■
The large-signal S-parameters from the .PRINT statement are written to
a .printls# file.
■
The small-signal S-parameters from the .PRINT statement are written to
a .printss# file.
■
The large-signal S-parameters from the .PROBE statement are written to
a .ls# file.
■
The small-signal S-parameters from the .PROBE statement are written to
a .ss# file.
■
The extracted large- and small-signal S and noise parameters are written to
a .p2d file.
The large-signal and small-signal S-parameters from the .PROBE statement
are viewable in Custom WaveView.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 193

N-2017.12
Chapter 8: Large Signal S-parameter Analyses
Large-Signal S-parameter (HBLSP) Analysis

HBLSP Analysis Limitations

The HBLSP analysis has these known limitations:
■
Power-dependent S-parameter extraction is a 2-port analysis only. Multiport
power-dependent S-parameters are not currently supported.
■
The intermodulation data block (IMTDATA) in the .p2d file is not supported.
■
The internal impedance of the port element can only be a real value.
Complex impedance values are not supported.

194 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 9
Envelope Analysis
9

Describes how to use envelope simulation.

The following topics are presented in this chapter:

■
Envelope Simulation
■
Envelope Analysis Commands
■ Output Syntax
■
Envelope Output Data File Format

Envelope Simulation
Envelope simulation combines the features of time- and frequency-domain
analysis. Harmonic Balance (HB) solves for a static set of phasors for all the
circuit state variables, as shown in this equation:

Equation 40

N
v  t  = a0 +   ai cos i t + bi sin i t 
i=1
In contrast, envelope analysis finds a dynamic, time-dependent set of phasors,
as this equation shows:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 195

N-2017.12
Chapter 9: Envelope Analysis
Envelope Analysis Commands

Equation 41

N
v  t  = a 0  ˆt  + ˆ ˆ
  ai  t  cos i t + bi  t  sin i t 
i=1
In envelope simulation, each signal is described by the evolving spectrum.
Envelope analysis is generally used on circuits excited by signals with
significantly different timescales. An HB simulation is performed at each point in
time of the slower-moving ( ˆt ) timescale. In this way, for example, a 2-tone HB
simulation can be converted into a series of related 1-tone simulations where
the transient analysis proceeds on the ( ˆt ) timescale, and 1-tone HB simulations
are performed with the higher frequency tone as the fundamental frequency.
In HSPICE advanced analog analyses, any voltage or current source identified
as a HB source either in a V or I element statement, or by an .OPTION
TRANFORHB command, is used for HB simulations at each point in ˆt time. All
other sources are associated with the transient timescale. Also, the input
waveforms can be represented in the frequency domain as RF carriers
modulated by an envelope by identifying a VMRF signal source in a V or I
element statement. The amplitude and phase values of the sampled envelope
are used as the input signal for HB analysis.
Some typical applications for envelope simulation are amplifier spectral
regrowth, adjacent channel power ration (ACPR), and oscillator startup and
shutdown analyses.

Envelope Analysis Commands

This section describes those commands specific to envelope analysis. These
commands are:
■ Standard envelope simulation (.ENV)
■
Oscillator simulation, both startup and shutdown (.ENVOSC)
■
Envelope Fast Fourier Transform (.ENVFFT)

196 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 9: Envelope Analysis
Envelope Analysis Commands

The following sections discuss these topics:

■
Nonautonomous Form
■
Oscillator Analysis Form
■
Fast Fourier Transform Form

Nonautonomous Form
.ENV TONES=f1 [f2...fn] NHARMS=h1 [h2...hn]
+ ENV_STEP=tstep ENV_STOP=tstop

Parameter Description

TONES Carrier frequencies, in hertz.

NHARMS Number of harmonics.

ENV_STEP Envelope step size, in seconds.

ENV_STOP Envelope stop time, in seconds.

Description
You use the .ENV command to do standard envelope simulation. The
simulation proceeds just as it does in a standard transient simulation, starting
at time=0 and continuing until time=env_stop. An HB analysis is performed
at each step in time. You can use Backward-Euler (BE), trapezoidal (TRAP), or
level-2 Gear (GEAR) integration.
Recommended option settings are:
■ For BE integration, set .OPTION SIM_ORDER=1.
■
For TRAP, set .OPTION SIM_ORDER=2 (default) METHOD=TRAP (default).
■
For GEAR, set .OPTION SIM_ORDER=2 (default) METHOD=GEAR.
Example
.env tones=1e9 nharms=6 env_step=10n env_stop=1u

Oscillator Analysis Form

.ENVOSC TONE=f1 NHARMS=h1 ENV_STEP=tstep ENV_STOP=tstop

HSPICE® User Guide: Advanced Analog Simulation and Analysis 197

N-2017.12
Chapter 9: Envelope Analysis
Envelope Analysis Commands

+ PROBENODE=n1,n2,vosc <FSPTS=num, min, max>

Parameter Description

TONE Carrier frequencies, in hertz.

NHARMS Number of harmonics.

ENV_STEP Envelope step size, in seconds.

ENV_STOP Envelope stop time, in seconds.

PROBENODE Defines the nodes used for oscillator conditions and the initial probe voltage value.

FSPTS Specifies the frequency search points used in the initial small-signal frequency
search. Usage depends on the oscillator type.

Description
You use the .ENVOSC command to do envelope simulation for oscillator startup
or shutdown.
Oscillator startup or shutdown analysis with this command must be helped
along by converting a bias source from a DC description to a PWL description
that either:
■ Starts at a low value that supports oscillation and ramps up to a final value
(startup simulation)
■
Starts at the DC value and ramps down to zero (shutdown simulation).
In addition to solving for the state variables at each envelope time point, the
.ENVOSC command also solves for the frequency. This command is intended to
be applied to high-Q oscillators that take a long time to reach steady-state. For
these circuits, the standard transient analysis is too costly. Low-Q oscillators,
such as typical ring oscillators, are more efficiently simulated with standard
transient analysis.
Example
.envosc tone=250Meg nharms=10 env_step=20n env_stop=10u
+ probenode=v5,0,1.25

Fast Fourier Transform Form

.ENVFFT output_var [NP=val] [FORMAT=keyword]

198 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 9: Envelope Analysis
Output Syntax

+[<WINDOW=keyword] [ALFA=val]

Parameter Description

output_var Any valid output variable.

NP The number of points to use in the FFT analysis. NP must be a power of 2. If not
a power of 2, then it is automatically adjusted to the closest higher number that is
a power of 2. The default is 1024.

FORMAT Specifies the output format:

NORM= normalized magnitude

UNORM=unnormalized magnitude (default)

WINDOW Specifies the window type to use:

RECT=simple rectangular truncation window (default)

BART=Bartlett (triangular) window
HANN=Hanning window
HAMM=Hamming window
BLACK=Blackman window
HARRIS=Blackman-Harris window
GAUSS=Gaussian window
KAISER=Kaiser-Bessel window

ALFA Controls the highest side-lobe level and bandwidth for GAUSS and KAISER
windows. The default is 3.0.

Description
You use the .ENVFFT command to perform Fast fourier Transform (FFT) on the
envelope output. This command is similar to the .FFT command. The only
difference is that the transformation is performed on real data with the .FFT
command, and with the .ENVFFT command, the data being transformed is
complex. You usually want to do this for a specific harmonic of a voltage,
current, or power signal.
Example
.envfft v(out)[1]

Output Syntax
The results from envelope simulation can be made available through the
.PRINT, .PROBE, and .MEASURE commands. This section describes the basic
syntax you can use for this purpose.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 199

N-2017.12
Chapter 9: Envelope Analysis
Envelope Output Data File Format

.PRINT or .PROBE
You can print or probe the envelope simulation results by using the following
commands:
.PRINT ENV ov1 <ov2... >
.PROBE ENV ov1 <ov2... >
Where ov1... are the output variables to print or probe.

.MEASURE
In HSPICE advanced analog analyses, the independent variable for envelope
simulation is the first tone. Otherwise, except for the analysis type, the
.MEASURE statement syntax is the same as the syntax for HB; for example,
.MEASURE ENV result ...

Envelope Output Data File Format

The results of envelope simulations are written to *.ev# data files by the
.PROBE statement. The format of an *.ev# data file is equivalent to an *.hb#
data file with the addition of one fundamental parameter sweep that represents
the slowly-varying time-envelope variation ˆt of the Fourier coefficients and
frequencies. You can recognize this swept parameter in the *.ev# file by the
keyword env_time.
Each row in the tabulated data of an *.ev# file includes values for identifying
frequency information, the complex data for the output variables, and
information on the envelope time sweep. For example, the header for a data file
dump for output variables v(in) and v(out), that follow a 2-tone envelope
analysis, have entries for:
hertz v(in) v(out) n0 f0 n1 f1 sweep env_time $&%#

Which result in data blocks with floating point values following:

200 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 9: Envelope Analysis
Envelope Output Data File Format

env_time[0]
f[0] a[0]{v(in)} b[0] {v(in)} a[0] {v(out)} b[0] {v(out)} n0
f0 n1 f1
f[1] a[1]{v(in)} b[1] {v(in)} a[1] {v(out)} b[1] {v(out)} n0
f0 n1 f1
...
f[N] a[N]{v(in)} b[N] {v(in)} a[N] {v(out)} b[N] {v(out)} n0
f0 n1 f1

env_time[1]
f[0] a[0]{v(in)} b[0] {v(in)} a[0] {v(out)} b[0] {v(out)} n0
f0 n1 f1
f[1] a[1]{v(in)} b[1] {v(in)} a[1] {v(out)} b[1] {v(out)} n0
f0 n1 f1
...
f[N] a[N]{v(in)} b[N] {v(in)} a[N] {v(out)} b[N] {v(out)} n0
f0 n1 f1

...

env_time[M-1]
f[0] a[0]{v(in)} b[0] {v(in)} a[0] {v(out)} b[0] {v(out)} n0
f0 n1 f1
f[1] a[1]{v(in)} b[1] {v(in)} a[1] {v(out)} b[1] {v(out)} n0
f0 n1 f1
...
f[N] a[N]{v(in)} b[N] {v(in)} a[N] {v(out)} b[N] {v(out)} n0
f0 n1 f1

Where there are M data blocks corresponding to M envelope time points, with
each block containing N+1 rows for the frequency data. The units for the
env_time sweep are seconds.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 201

N-2017.12
Chapter 9: Envelope Analysis
Envelope Output Data File Format

202 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Part 3: Large Signal Noise Analyses

Note: For .AC-related .NOISE analysis, see AC Small-Signal and

Noise Analysis in the HSPICE User Guide: Basic Simulation and
Analysis.
This Part contains the following chapters/topics:
■
Chapter 10, Simulation of Random Noise
■ Chapter 11, Transient Noise Analysis
■
Chapter 12, Periodic Time-Dependent Noise Analysis (.PTD-
NOISE)

HSPICE® User Guide: Advanced Analog Simulation and Analysis 203

N-2017.12
204 HSPICE® User Guide: Advanced Analog Simulation and Analysis
N-2017.12
 10
Simulation of Random Noise
10

Describes the characteristics of random signals, types of noise, component

noise models, and noise simulation in HSPICE.

HSPICE ships several examples for your use; see Applications of General
Interest Examples in the HSPICE User Guide: Basic Simulation and Analysis
for paths to the demo files.

Note: For .AC-related .NOISE analysis, see AC Small-Signal and

Noise Analysis in the HSPICE User Guide: Basic Simulation and
Analysis.

This chapter presents the following topics:

■
Introduction to Noise Sources
■ Characteristics of Random Signals
■
Noise Types
■
Component Noise Models and HSPICE Noise Simulation

Introduction to Noise Sources

One of the essential techniques in designing an analog circuit is to be able to
identify major noise sources in the circuit and know how to minimize their effect
to an acceptable level. Noise can be loosely defined as a disturbance signal
with random amplitude, which is usually an unwanted phenomenon in a circuit.
It is important to limit the power of noise to a low level relative to that of the
signal; otherwise, the signal can be greatly distorted or completely
indistinguishable. They rely on simulation tools to help them identify the
dominant sources of noise in the circuit, measure the noise levels (in power,

HSPICE® User Guide: Advanced Analog Simulation and Analysis 205

N-2017.12
Chapter 10: Simulation of Random Noise
Characteristics of Random Signals

current, or voltage), and eventually make changes in the design to reduce the
noise levels, if necessary.
Two types of noise are commonly defined: inherent and interference.
■
Inherent noise is the noise created by the elements internal to the circuit
(resistors, diodes, and transistors), caused by movement of electrons in the
circuit.
■
Interference noise refers to the unwanted signals from the environment
absorbed by the circuit.
Eliminating interfering noise can be done by different methods of signal
shielding or code modulation so that the signal can be differentiated from the
noise. Inherent noise, however, can only be affected by changes in the circuit
topology and the bias currents.
This chapter discusses only inherent noise and the HSPICE simulations related
to it. Whenever the term “noise” is used, it refers to “inherent noise”.
Presented first in the following sections are some theoretical definitions for
noise sources in an analog circuit. These are followed by a discussion on how
HSPICE performs a noise simulation and how to interpret outputs from a noise
simulation.

Characteristics of Random Signals

The following sections describe noise sources in the time and frequency
domains, and methods to deal with the multitude of noise sources in a circuit:
■
Probability Distribution Function versus Power Spectral Density
■
Multiple Noise Sources in a Circuit
■
Summary

Probability Distribution Function versus Power Spectral

Density
Noise is a signal with random amplitude. Figure 19 illustrates a sample noise
signal in the time domain. The vertical axis could be either a branch current or
a node voltage.

206 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Characteristics of Random Signals

Figure 19 Typical noise signal (voltage or current) in the time domain

To simulate a noise source in the time domain, you have to know the Probability
Distribution Function (PDF) of that signal and then create a voltage or current
source that produces a random output signal conforming to the PDF. Figure 20
demonstrates a sample PDF for a random signal with a Gaussian distribution.
Extracting PDF parameters from the circuit components is not straightforward
and therefore not common in analyzing noise.

PDF

Signal Amplitude

Figure 20 PDF for a random signal with Gaussian probability distribution

However, noise can be easily modeled and measured in the frequency domain.
The noise models in the frequency domain also relate directly to the
component and circuit characteristics. This is the reason noise analyses are
usually done in the frequency domain.
While we cannot obtain a deterministic value for the noise amplitude in the time
domain because it is a random variable, it is quite possible to determine the
power of the noise if the noise is bandwidth limited. In other words, it is true that
the amplitude of a noise signal changes randomly in the time domain. However,
its power, as defined by Equation 42 is a fixed and finite value if and only if the

HSPICE® User Guide: Advanced Analog Simulation and Analysis 207

N-2017.12
Chapter 10: Simulation of Random Noise
Characteristics of Random Signals

signal is bandwidth limited.

Definition of Power for a noise voltage (a) and current (b), normalized for a 1-
ohm load:

Equation 42
T2
P = lim T    1  T 
2
v  t  dt Eq. a
–T  2
T2
P = lim T    1  T 
2
i  t  dt Eq. b
–T  2
The power of a signal can also be described in the frequency domain; by
means of a Power Spectral Density (PSD). A PSD identifies the portion of
signal power present as a function of signal frequency. Figure 21 illustrates two
PSD graphs.
Graph A resembles the PSD of a typical data signal in high speed and RF
communication; a signal with a limited bandwidth. It can be seen from the
graph that the composing frequencies of the signal are between DC (0 Hz) and
200 MHz, with the bulk of power being concentrated around 100 MHz.
Graph B shows a signal with a flat power distribution; meaning that the power is
equally distributed across all frequencies. Such a signal is referred to as the
“white noise”. Resistor thermal noise possesses such a PSD.

PSD PSD

¶ ¶
100 MHz 200 MHz
A B
Signal with the bulk of power Signal with power equally
concentrated around 100 MHz distributed across all frequencies

Figure 21 Sample PSD graphs

To obtain the total power of a signal within a certain bandwidth, you must
calculate the area under the PSD curve within the given frequency range, as
defined in Equation 43.

208 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Characteristics of Random Signals

Definition of power based in the frequency domain, normalized for a 1 ohm

load:

Equation 43
f_2
P = f_1 PSD  f df
An example of power calculation in the frequency domain for a noise signal is
shown in Figure 22. In this example, the PSD is a flat curve with a value of
5 x 10-20 V2/Hz.

PSD
5 x 10-20

¶
500 MHz 800 MHz

Figure 22 Sample PSD graphs

The power of the noise between the frequencies 500 MHz and 800 MHz is
equal to the shaded area under the curve and is given by:
800M
500M 5  10
– 20 – 20 – 11 2
Pn = df =  800M – 500M 5  10 = 1.5  10 V

The root mean square (RMS) value of the noise signal is given by Equation 44.
The RMS value of a signal is defined as the DC value that would render the
same power as the signal in question. For the general sinusoidal signal
Acos(t), the RMS value is A  2 .
The RMS value of a signal is the root of the 1 ohm normalized signal power:

Equation 44
v rms = P
Which means that the RMS voltage value of the noise signal in the above
example is:
– 11
v rms = 1.5  10 = 5.48V

HSPICE® User Guide: Advanced Analog Simulation and Analysis 209

N-2017.12
Chapter 10: Simulation of Random Noise
Characteristics of Random Signals

This means that the power of the noise is identical to that of a sinusoidal signal
with a RMS value of 5.48 uV.

Multiple Noise Sources in a Circuit

In the usual case of dealing with a multitude of noise sources in a circuit, it is
important to know how to add up the effects of noise to obtain the accumulated
noise value.
If the noise sources are completely independent (that is, un-correlated), such
as the noises generated by two separate resistors, their powers add up to their
RMS values. Figure 23 shows how two independent noise sources, in this case
both voltage sources, add up.

*
Figure 23 Noise summation: independent noise signals add in squared values

It might not be evident from the equation for VnTotal in Figure 23 that when there
are several noise sources in a circuit, only reducing the biggest noise
contributors will reduce the total noise significantly. Reducing the smaller
sources, even by large amounts does not have a considerable effect on the
overall noise value because the power of two of the signals magnifies the
dominance of bigger sources over smaller ones.
To clarify this point, consider the case where two noise sources are present,
with RMS values of v1 = 2 uV and v2 = 9 uV. The total noise is given by:

2 2
v nTotal =  2  +  9  = 9.21V

210 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Noise Types

Now if in an attempt to lower the total noise the designer of the circuit cuts the 2
uV source by 50%, the total noise will be:
2 2
v nTotal =  1  +  9  = 9.06V
Now, if instead we cut the 9 uV source by only 10%, the total noise will be:
2 2
v nTotal =  2  +  8.1  = 8.34V
This emphasizes the fact that to reduce the total noise, we should first
concentrate on the strongest source of noise.

Summary
Important points from this section are:
■
Inherent noise is best modeled in the frequency domain by its PSD.
■
The power of noise depends on the bandwidth of the system and is defined
as the area under the PSD curve between two given frequencies.
■
When multiple noise sources are present, the square of their voltage (or
current) values add up to produce the total noise.
■
When multiple sources of noise are present in a circuit, the most effective
way of reducing the total noise is to focus on reducing the most dominant
noise contributor.

Noise Types
The following sections identify and describe three common types of noise in
analog circuits:
■
Thermal Noise
■ Flicker Noise
■
Shot Noise
■
Summary

HSPICE® User Guide: Advanced Analog Simulation and Analysis 211

N-2017.12
Chapter 10: Simulation of Random Noise
Noise Types

Thermal Noise
Thermal noise is generated by random movements of electrons in a resistive
conductor. Any circuit element with resistive characteristics, whether a resistor
or the base junction series resistor in a BJT’s small signal model, disseminates
thermal noise. The models for thermal noise voltage source and their
equivalent current source are shown in Figure 24.

R * vn

R in

A B C
Resistor Resistor modeled as a noiseless Resistor modeled as a noiseless
resistor and a series noise voltage resistor and a shunt current source

Figure 24 Thermal noise voltage sources and their equivalent circuits

The voltage and current PSD functions for a resistor thermal noise are:
2 2
V n  f  = 4kTR v  Hz
2 2
I n  f  = 4kT  R A  Hz
where:
k is Boltzmann’s constant equal to 1.38 x 10-23.
T is the temperature in Kelvin degrees: 1 °K = °C + 273. For a room
temperature of 25 °C (88 °F), the default temperature setting in HSPICE, T will
be 298 °K.
R is the resistance in ohms.
It should be noted that the two voltage and current PSD models are equivalent,
they both produce the same amount of open circuit voltage or short circuit
current and can be used interchangeably during noise analysis.
Notice that the thermal noise voltage PSD is proportional to the resistance
while the current PSD is inversely proportional to the resistance. By looking at

212 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Noise Types

the PSD models, a general conclusion can be made: the presence of bigger
resistors in a circuit usually translates to a bigger thermal noise voltage.
The PSD models have been explicitly described as functions of frequency to re-
emphasize the fact that the PSD is only meaningful in the frequency domain
context.
The above models suggest that PSDs are flat across all frequencies. This
assumption will no longer be true for frequencies above a few hundred
gigahertz.
Example 1
What will be the RMS voltage value of the thermal noise generated by a 10
kohm resistor operating in the 1- to 1.2-GHz frequency range? Assuming a
room temperature of 25 °C or 298 °K, the PSD function will be (see also
Figure 25):

2 – 23 – 16 2
V n  f  = 4kTR = 4  1.38  10   298   10K  = 1.64  10 V  Hz
2
 PSD  f df =   1.64  10
2 – 16 – 16 –8
v nrms =  df =  1.64  10    0.2G  = 3.28  10 V

–8
v nrms = 3.28  10 = 181.11V

Power = 3.28x10-8
PSD

* vn = 181.11mv
1.64x10-16

10 KW


1 GHz 1.2 GHz

A B
Noise Voltage Model PSD Model

Figure 25 Noise models for 10 kohm resistor

HSPICE® User Guide: Advanced Analog Simulation and Analysis 213

N-2017.12
Chapter 10: Simulation of Random Noise
Noise Types

Flicker Noise
Flicker noise is the noise generated by the fluctuations in the average current
travel time in a conductor. As electrons take different random paths to get from
one end of a resistor to the other, at any given time, the average current will
experience different resistance along the way, leading to fluctuations in current,
hence the generation of flicker noise. Figure 26 illustrates how electrons can fall
into paths with different lengths while flowing through a resistor.

Electrons taking a long path to get through the resistor

Electrons taking a short path to get through the resistor

Figure 26 Paths for current through a resistor

By shrinking the cross-section area of the conducting material, there will be

smaller differences between different current paths and the flicker noise will
shrink. That is why thin film resistors have a much smaller flicker noise than
carbon rod resistors, which have a much larger conducting cross-section.
In addition, at high frequencies, the current tends to travel through the outer
surface of the conductor, a phenomenon know as “skin effect”, effectively
shrinking the conducting cross-section to a very thin layer. This results in
smaller differences in the current path lengths and consequently smaller flicker
noise. Because of stronger flicker noise effects at higher frequencies, flicker
noise is inversely proportional to the frequency. That is why it is also called “1/f
noise”. Flicker noise is only significant at lower frequency values, but it can
have a major impact on nonlinear circuits at RF frequencies.
The major sources of flicker noise in analog circuits are semiconductor
components. Resistors can be manufactured to produce only small a amount of
flicker noise and are not usually taken into account for the calculation of flicker
noise in a circuit.

214 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Noise Types

Flicker noise is usually represented as a current PSD for resistors and

semiconductors with this equation:
2 2
In  f  = K f I  f A  Hz
Where,
I is the current through the resistor.
f is frequency in hertz.
Kf is a constant which depends on the characteristics and geometry of the
conductor.
The PSD for flicker noise is depicted in Figure 27.

PSD

Figure 27 PSD for flicker noise

HSPICE uses more general models for CMOS flicker noise PSDs. One of the
models used for a CMOS transistor is the SPICE2/Berkeley noise model:
2 A 2 E 2
I n  f  = K f I f   C ox L f f  A  Hz
where:
Kf is the flicker noise parameter.
Af is the current exponent.
Cox is the Gate oxide capacitance.
L is the effective length of the transistor.
Ef is the frequency exponent.
Another model used by HSPICE for flicker noise calculation is the BSIM model,
which looks very similar to the above equation. It is important to note that you
usually must not be concerned with the value settings of these parameters.
They must be defined and set to proper values in the transistor models defined
in the technology library.
The combined PSD for flicker and thermal noise is shown in Figure 28 on
page 216.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 215

N-2017.12
Chapter 10: Simulation of Random Noise
Noise Types

PSD

Thermal noise
floor

Figure 28 PSD of combined thermal and flicker noise

Note that the flicker noise is the dominant source of noise at lower frequencies
and as its magnitude shrinks at higher frequencies, the thermal noise will
become the dominant source of noise. Thermal noise is also referred to as “the
noise floor” because that is the minimum possible amount of noise in a circuit.

Shot Noise
Shot noise is only generated by semiconductor elements and is caused by
random passage of electrons and holes across a potential barrier, such as a
P-N junction. Shot noise is often represented by a current PSD, known as the
“Schottky formula”:
2 2
I n  f  = 2qI A  Hz
Where,
q = 1.6 x 10-19 C is the charge of an electron.
I is the bias current of the junction and can be Ic, collector bias currents for a
BJT transistor.
Shot noise PSD curves are flat across all frequencies.

216 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

Summary
The three major types of noise in an analog circuit are:
■
Thermal noise, also known as white noise, is generated by the resistors in
the circuit. Thermal noise is a function of conductor resistance.
■
Flicker noise, also called 1/f noise, is mainly generated by transistors in a
circuit. It is a function of the component geometry and its magnitude drops
as frequency increases.
■
Shot noise is caused by bias currents in the base and collector of BJT
transistors.

Component Noise Models and HSPICE Noise

Simulation
The following sections describes how HSPICE models noisy elements,
calculates the total output noise PSD, and obtains the total output noise voltage
PSD, and how the RMS output noise voltage can be deduced from the total
output noise PSD.
■
Element Noise Models
■
HSPICE Noise Simulation
■
Summary

Element Noise Models

Each resistor, diode, and transistor in a circuit generates at least some type of
inherent noise. HSPICE models the noise generating elements in a circuit as a
noiseless element combined with a noise current or voltage source with a
known PSD. Figure 29 on page 218 includes the listing of the four noise
generating elements in analog circuits and their equivalent noise models.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 217

N-2017.12
Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

Thermal noise
voltage

Channel resistance

* V2 (f)

V2 (f)
thermal noise voltage

I2 (f)
R
*
Flicker noise
A voltage B
Resistor and Diode noise model MOS noise model

Thermal noise voltage for Shot noise current due

Base-Emitter series resistance to Collector bias current

V2 (f)

I2 (f) * I2 (f)

Combined flicker noise and

Base bias current shot noise
C
BJT noise model

Figure 29 Noise generating circuit elements and their noise models

The noise model for a diode is identical to that of a resistor with a resistance of:
R d = kT  qI d
Where,
Rd is the equivalent resistance of the diode.
k is Boltzmann’s constant.
q is the charge of an electron.
Id is the bias current of the diode.

HSPICE Noise Simulation

To perform a noise analysis in HSPICE, a .NOISE statement along with an .AC
statement must be used. The .NOISE syntax for the noise analysis is:

218 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

.NOISE outvar1 [outvar2 outvar3...] [vin] [interval]

+ [portnoise[=1|0]]
+ [listckt [=1|0]]
+ [listfreq=frequencies|none|all]
+ [listcount=num] [listfloor=val]
+ [listsources=1|0|yes|no]
+ [listtype=1|0]
The frequency points at which the noise calculations are performed are the
same points defined by the .AC statement. The noise calculations for each
frequency point will be output to the listing file.
The following sections discuss these topics:
■
RX Transfer Function
■
Example

RX Transfer Function
RX is the transfer function from the noise source to the circuit output. Since the
noise sources for built-in devices are all current sources with PSDs (power
spectral densities) in A2/Hz, and the output is usually a voltage (specified in
the .NOISE statements), RX is a transimpedance (V/A).
The total output noise voltage is the integrated output noise. The output noise
at a given frequency is a PSD in V2/Hz. This can be integrated over the
frequency range (using the .AC command) to get a total output noise in V2. If
you take the square root, you get the total output noise in V.

Example
For example (see Example 30 on page 220), take a simple common source
NMOS amplifier to show how HSPICE calculates the output noise. In this
example, the output port element (P2) is used as a pull-up resistor in the circuit.
For the path to the demo file, noise_app_orig.sp, see Applications of
General Interest Examples in HSPICE User Guide: Basic Simulation and
Analysis. The full path to this example is $installdir/demo/hspice/
apps/noise_app_orig.sp.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 219

N-2017.12
Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

vdd

* A Common Source NMOS amplifier

.options list post
Port 2 .model n_tran nmos level=49 version=3.22 AF=.826 KF=4e-29

Output vdd vdd 0 DC=5

p1 in 0 port=1 ac=0.1 dc=2.1 z0=50
Rs d p2 out vdd port=2 z0=20k
rs in g1 50
s m1 out g1 0 0 n_tran l=1.5u w=40u
Input .ac dec 10 10Meg 10G
.noise v(out) p1 1
Port 1
.print ac v(out) onoise

.end

Figure 30 A common source NMOS amplifier and its netlist

The first step in the noise analysis is to set all the signal voltage and current
sources to 0. The equivalent circuit for our example, after setting vdd=0, is
shown in Figure 31.

Output

Rs d
Input

50 s

Figure 31 Equivalent circuit after independent V and I sources are set 0

HSPICE models each resistor, diode, and transistor with its noise model, and
then calculates the output voltage resulting from the noise signal, one element
at a time.
To start with, it replaces Rs with its noise model, as shown in Figure 32, and
calculates the PSD of the noise voltage as seen at the output port. HSPICE
reports an output voltage PSD of 85.4443x10-18 V2/Hz, caused by the thermal

220 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

noise model of Rs. The value rx shown is used to obtain the voltage transfer
function from the Rs noise source to the output port:
output_voltage
Voltage Transfer Function = rx = ------------------------------------------------------
element_noise_current
HSPICE uses rx for its internal calculations and you need not pay any particular
attention to it.

2 – 18
V n  f  = 85.443  10

frequency = 100.0000x hz
Output

Rs d *** resistor squared noise voltages (sq v/hz)

Input
+
–

50 s
element 0:rs
Vn (f)

total 85.4443a
rx 509.3796

Figure 32 Circuit noise model for Rs and analysis output at 100 MHz

The circuit for calculating the PSD of the output noise generated by the NMOS
is given in Figure 33.

2 – 15
V n  f  = 3.4691  10
frequency = 100.0000x hz

** mosfet squared noise voltages (sq v/hz)

Output

Rs Vn (f) d element 0:m1

Input In(f)
+ rd 0.
–

50 s rs 0.
id 3.4544f
rx 18.4636k
fn 14.7854a
total 3.4691f

Figure 33 Circuit noise model for NMOS and analysis output

HSPICE® User Guide: Advanced Analog Simulation and Analysis 221

N-2017.12
Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

The total PSD at 100 MHz will be the sum of all individual PSDs (the square of
noise signals addition):

2 – 18 – 15 – 15 2
V nTotal  f = 100MHz  = 85.4443  10 + 3.4691  10 = 3.5546  10 V  Hz
– 15 –9
V nTotal  f = 100MHz  = 3.5546  10 = 59.6204  10 V  Hz
It is common to represent the total noise generated by a circuit as the
equivalent input noise (otherwise called as the input referred noise). The input
referred noise voltage/current is the voltage/current that, if applied at the input
of the noise-free circuit, would have generated the same output voltage as the
total output noise voltage.
To calculate the equivalent input noise signal, you have to obtain the transfer
function from the input to the output. For the example circuit above, the transfer
function will be:
Output Voltage V  out 
TF = ------------------------------------- = -----------------
Input Voltage VS
In the case where the input source is a current source instead of a voltage
source, the TF would be the ratio of the output voltage over the input current.
The equivalent input noise voltage (or current in case of the input current
source) is given by:
Total Output Noise Voltage
Input Referred Noise = ------------------------------------------------------------------
TF
The input referred noise is useful to calculate the circuit's “noise figure,” which
is a measure of how much a circuit is generating inherent noise. A large noise
figure indicates a high level of noise generation by the circuit. Noise figure is
defined as:
Noise Figure = 10  log   Noise Factor 
Where Noise Factor is given by:
Input Referred Inherent Noise Power
Noise Factor = 1 + -----------------------------------------------------------------------------------------
Power of External Noise at the Input
Example 34 on page 223 shows the HSPICE noise analysis summary output
for the above circuit.

222 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

2
Vn  f = 100MHz 

V n  f = 100MHz 
**** total output noise voltage = 3.5546f sq v/hz

= 59.6204n v/rt hz

TF
V n  f = 100MHz   TF
transfer function value:
v(out)/vs = 10.1876
equivalent input noise at vs = 5.8523n /rt hz

**** the results of the sqrt of integral (v**2 / freq)

from fstart up to 100.0000x hz. using more freq points
results in more accurate total noise values. 100Meg
10Meg
2
V rms = V n  f  df

**** total output noise voltage = 570.7076u volts

V rms   TF 

**** total equivalent input noise = 55.7041u

Figure 34 Noise summary for noise performed at 100 MHz

Summary
To summarize,
■
HSPICE model elements as noiseless elements connected to noise-
generating voltage and current sources.
■
To calculate the total output noise PSD, set all signal and supply sources to
zero. The simulator then replaces, one-by-one, the noise-generating
elements with their equivalent noise models and calculates the noise

HSPICE® User Guide: Advanced Analog Simulation and Analysis 223

N-2017.12
Chapter 10: Simulation of Random Noise
Component Noise Models and HSPICE Noise Simulation

voltage PSD at the output caused by the element's noise source (similar to
an .AC analysis). It then repeats the same process for the next noise-
generating element.
■
Once the output noise PSDs are calculated for all elements, HSPICE adds
the PSDs together to obtain the total output noise voltage PSD.
■
The RMS output noise voltage and the input referred noise can be deduced
from the total output noise PSD.

224 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 11
Transient Noise Analysis
11

Describes the HSPICE solutions to perform transient noise analysis and

compute noise statistics and their variation over time for circuits driven with non-
periodic waveforms.

Transient noise analysis shows the effect of noise on the signal magnitude. It is
also useful to see how noise affects the timing of the signal. Transient noise
simulations include statistically accurate noise source contributions from all
devices. The noise source frequency response is adjustable. From the transient
noise analysis results, you can extract and measure phase noise and jitter. This
capability is especially useful in characterizing phase locked loops (PLL). The
two jitter measurements are time interval error (TIE) and autocorrelation
function. TIE measures the time-shift behavior relative to a reference signal.
The autocorrelation function tracks the relative time-shift behavior of the signal.
This chapter describes two approaches:
■ Monte Carlo (default), where HSPICE uses random signal sources to
predict the statistical characteristics of the circuit performance due to device
noise. (See Monte Carlo Noise Analysis on page 229.)
■ Stochastic Differential Equation (SDE), for advanced users, which makes a
direct prediction of the actual statistics of the output waveforms. (See
Stochastic Differential Equation Analysis on page 245.)
HSPICE includes several different algorithms for understanding circuit behavior
based on the noise generated internally by electronic devices and thermal
noise. PHASENOISE analysis computes the effects that noisy elements have
on the output spectrum of oscillators. HBNOISE and SNNOISE analyses
compute the small-signal variations that noise can create under large-signal
steady-state operating conditions (see Multitone Harmonic Balance Noise
(.HBNOISE) or Shooting Newton Noise Analysis (.SNNOISE). Periodic time-
domain noise (PTDNOISE) analysis computes the noise statistics of a periodic
signal, and how those statistics vary with time over the period of the steady-

HSPICE® User Guide: Advanced Analog Simulation and Analysis 225

N-2017.12
Chapter 11: Transient Noise Analysis
Overview of HSPICE Transient Noise Analysis

state signal (see Periodic Time-Dependent Noise Analysis (.PTDNOISE). What

makes transient noise analysis unique compared to these other approaches is
that its noise analysis is a complement to the fully nonlinear dynamic time-
domain .TRAN simulation of HSPICE.
For additional information and graphical illustration of the HSPICE transient
noise solution, see the webinar link: Jitter Analysis Using HSPICE Transient
Noise Techniques at http://www.hspice.com and https://www.synopsys.com/
news/pubs/snug/sanjose11/ma6_tutorial_HSPICE_trannoise.pdf (requires your
Synopsys SolvNet user ID and password).

Note: For .AC-related .NOISE analysis, see AC Small-Signal and

Noise Analysis in the HSPICE User Guide: Basic Simulation and
Analysis.

The following sections discuss these topics:

■
Overview of HSPICE Transient Noise Analysis
■
Modeling Frequency-Dependent Noise Sources
■
Monte Carlo Noise Analysis
■
Setting up a Monte Carlo Transient Noise Analysis
■
Controlling Noise
■
Setting Noise Bandwidth
■
TIE and Spectrum Jitter Measurements from Transient Noise Results
■
Phase Noise and Jitter from Transient Noise
■
Power Spectral Density from Transient Noise
■
Stochastic Differential Equation Analysis
■ Error Handling, Error Recovery, Status Reporting
■
References

Overview of HSPICE Transient Noise Analysis

A variety of noise measurements are desirable from circuit simulation. The
traditional SPICE .noise analysis provides a measurement of the RMS noise
voltage at an output node as a function of frequency. This RMS value is similar
to a measurement of the standard deviation of an equivalent Gaussian

226 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Overview of HSPICE Transient Noise Analysis

distribution of noise present at the output node of interest due to the

contributing random noise sources within the circuit. The .noise analysis is a
small-signal analysis that gives an output noise (onoise) value over the .ac
frequency range.
More advanced examples of noise measurements include phase noise and
timing jitter. Timing jitter, in particular, is a measurement of a clock or
oscillator's random noise over a time interval. Timing jitter represents the
standard deviation (or variance) of the timing uncertainty (that is, the random
drift of the clock edges) as a function of time. It is therefore a time-domain noise
measurement that HSPICE typically evaluates at each cycle or half-cycle. For
some clocks, oscillators, and PLLs, you can separate this measurement into
time-independent and time-dependent contributions written as:

Equation 45
2 2 -
    = -----------
2
 R  0  – R    
0
2
where     represents the time-dependent variance,  is the time advance,
and R     is the autocorrelation function that relates to the power spectrum of
phase variations as:

Equation 46

1 j 
R     = ------ S    e d
2
–

These special relationships allow the computation of timing jitter from the
results of .phasenoise analysis, since we can assume L  f   S   f  and
therefore derive the time-varying noise from the frequency-domain phase noise
simulation solution.
Other measurements of time-dependent or time-varying noise are also
desirable for circuits other than clocks and oscillators, and for situations other
than the steady-state operation. The purpose of such measurements is usually
similar: derive useful information on the time-varying statistical behavior of the
circuit due to its internal noise sources.
Transient noise analysis is a typical .TRAN simulation but with all random noise
sources within the circuit activated as contributing signal sources. The Monte
Carlo noise analysis is a transient noise simulation approach that uses
uncorrelated random signal sources for device noise in such a way that you can

HSPICE® User Guide: Advanced Analog Simulation and Analysis 227

N-2017.12
Chapter 11: Transient Noise Analysis
Modeling Frequency-Dependent Noise Sources

seed all noise signals uniquely. In addition, you can repeatedly, from run-to-run,
predict the statistical characteristics of the circuit performance due to device
noise. You can typically examine the resulting outputs by using histogram plots
to measure their statistical behavior.

Modeling Frequency-Dependent Noise Sources

Transient noise techniques require that you model noise sources in the time
domain. These techniques model device noise sources in terms of the standard
white Gaussian noise processes. You write the standard white Gaussian noise
process as   t  .
You calculate white noise source models in terms of intensity functions. For
example, if Equation 47 gives the (frequency dependent) noise from a time
varying conductance:

Equation 47
2
i n = 4kTG  t f
Then you can model in terms of intensity as the random time-domain current
given by:

Equation 48
jn = 2kTG  t   t 
In the case of flicker noise, it is necessary to create the 1/f power spectrum of
flicker noise sources by filtering a   t  process. You can accomplish this with
the following rational function transfer relationship:

Equation 49
N  s +  zi 
Hs =  --------------------
 s +  pi 
-
i =1

Since you can model all flicker sources over the same frequency range (that is,
bandwidth), you can use this same transfer function for all sources. Tests show
that the above fit is very accurate using one pole/zero per octave. This
modeling technique shows reasonable fits using one pole/zero for every two
octaves, or even with one pole/zero per decade. This fitting algorithm is
therefore frequency range-specific, which is why you use the parameters FMAX

228 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Monte Carlo Noise Analysis

and FMIN for specifying frequency ranges for transient noise analysis similar to
the modeling of the frequency-dependent S-element.
The TRANNOISE analysis includes a time-domain noise source for all noise
contributions within all devices. So, for example, if the transistor model in use
includes induced gate noise, so does .TRANNOISE. For example, if you use a
BSIM4 model and the model parameter set includes values for Induced Gate
Noise, then .TRANNOISE includes this noise in the simulation.

Monte Carlo Noise Analysis

The equations for Monte Carlo transient noise analysis define the Monte Carlo
approach.
The time-domain system of equations for transient analysis drives the Modified
Nodal Analysis (MNA) system described by the following vector state equation:

Equation 50
f  x· x t  = 0
Consider this the noiseless system that HSPICE solves for during a normal
transient analysis. You can consider the transient noise analysis aa a similar
analysis where we now inject noise from all device noise models to give the
modified equation:

Equation 51
f  x· x t  = – j n  t 
This system reflects the added noise sources and HSPICE solves it in the
same manner as transient analysis. However, if we consider each noise source
related to a white Gaussian noise function, we must create an ensemble of
waveforms for our unknown x(t) vector to predict the output statistics. This type
of simulation involves generating multiple uncorrelated noise source waveforms
for all noise sources, and then running multiple simulations in a Monte Carlo
fashion. This approach is called the Monte Carlo noise simulation. This method
cannot directly measure the statistics of output signals due to input noise (as
with .NOISE), but instead models noise sources as independent time-domain
stimuli. Generating statistical information in this approach requires running a
plurality of simulations over a variety of random noise-source sequences (to
create an ensemble of output waveforms) and then analyzing the statistics of
the ensemble using histograms of other plots. When the system behavior is
ergodic, it is possible to run a single very long duration simulation in order to

HSPICE® User Guide: Advanced Analog Simulation and Analysis 229

N-2017.12
Chapter 11: Transient Noise Analysis
Setting up a Monte Carlo Transient Noise Analysis

capture the statistical variations of the output signal over time (as with an eye-
diagram). Monte Carlo modeling of the noise sources uses a sum of sinusoids
with random phases [2], or random number generators with the appropriate
statistical distributions [3]. An advantage of the Monte Carlo approach is its
ability to capture very nonlinear noise behaviors. This is useful, for example,
when you know that the responses of circuits with noise have non-Gaussian
variations about their noiseless simulations.

Setting up a Monte Carlo Transient Noise Analysis

The transient noise analysis requires an accompanying .tran analysis which
determines the time-sampling, matrix solutions, and deterministic output
waveforms. The .TRANNOISE command activates transient noise and
computes the additional noise variables. This is consistent with how .NOISE
computes additional noise outputs when you add it to an .AC analysis.
The following sections discuss these topics:
■
Transient Noise and Monte Carlo Noise Analysis Techniques Available in
HSPICE
■
Monte Carlo Output Data
■
Monte Carlo Examples

Transient Noise and Monte Carlo Noise Analysis

Techniques Available in HSPICE
The four methods of using Monte Carlo analysis and transient noise analysis
are:
■
Traditional Monte Carlo with Transient Noise Disabled
■ Transient Noise Analysis without Monte Carlo (Single Sample)
■
Transient Noise Analysis with Monte Carlo
■
Both Transient Noise and Traditional Monte Carlo Enabled

Traditional Monte Carlo with Transient Noise Disabled

To use traditional Monte Carlo with transient noise disabled, specify a Monte
Carlo run with the .TRAN command.

230 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Setting up a Monte Carlo Transient Noise Analysis

For example:
.TRAN 10p 8u SWEEP MONTE=20 FIRSTRUN=2
* gives 20 Monte Carlo samples from 2-21
* no .trannoise

This method uses the standard SWEEP MONTE keywords to activate Monte
Carlo parameter sampling for GAUSS, AGAUSS, UNIF, and AUNIF defined
parameters. For this analysis, the transient noise is not included. The result of
the simulation is an ensemble of output waveforms that reflect no noise, but
only the traditional Monte Carlo parameter variations.

Transient Noise Analysis without Monte Carlo (Single Sample)

A single transient noise analysis includes statistically accurate time-varying
noise contributions for all noise sources present in the circuit, including SPICE
models. The bandwidth is adjustable for all noise-source frequency responses.
This analysis approach is the fastest possible based on single .TRAN analysis
and is the most effective for viewing statistical variations over time (ergodic
simulations).
To use transient noise analysis without Monte Carlo, specify a single sample
transient noise analysis.
For example:
.TRAN 10p 8u $.tran without SWEEP MONTE
.TRANNOISE v(out<0>) $ output of interest
+ fmin=1k fmax=1e8 $ sets noise bandwidth

When setting the noise bandwidth, FMIN is used to set the low-frequency flicker
noise limits and FMAX is used to set the maximum noise-source waveform
bandwidth and ensure Nyquist sampling. The seed values can be reset to re-
run simulations with uniquely different noise waveforms. The output waveforms
can be post-processed to measure the resulting noise effects.

Transient Noise Analysis with Monte Carlo

To use transient noise analysis with Monte Carlo, specify a multi-sample
transient noise analysis.
For example:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 231

N-2017.12
Chapter 11: Transient Noise Analysis
Setting up a Monte Carlo Transient Noise Analysis

.TRAN 10p 8u $.tran without SWEEP MONTE

.TRANNOISE v(out<0>) $ output of interest
+ METHOD=MC $ use random noise waveform
+ SAMPLES=20 $ 20 samples
+ SEED=2 $ begin with seed=2, ends at 21
+ fmin=1k fmax=1e8 $ sets noise bandwidth

For this analysis, traditional Monte Carlo parameter sampling is disabled and
only the nominal values are used for GAUSS, AGAUSS, UNIF, and AUNIF
defined parameters during the simulation.
The transient noise METHOD=MC keyword will inject random noise waveforms
for all noise sources present in the circuit, including SPICE models. The result
of the simulation is an ensemble of output waveforms that reflect the random
noise waveform injections. Each noise seed sample will have a unique
waveform. This approach is useful for characterizing statistics at specific time
events and the waveforms can be post-processed as a histogram.

Both Transient Noise and Traditional Monte Carlo Enabled

To use transient noise analysis with the traditional Monte Carlo analysis
enabled, specify both a traditional Monte Carlo run and a transient noise
analysis.
For example:
.TRAN 10p 8u SWEEP MONTE=20 FIRSTRUN=2
* sets Monte seed values from 2-21
.TRANNOISE v(out<0>) $ add .trannoise
+ METHOD=MC
+ SAMPLES=1 SEED=2 $ THIS IS IGNORED
* Monte Carlo seeds 2-21 defined in .TRAN are used
+ fmin=1k fmax=1e8 $ sets noise bandwidth

The Monte Carlo seeds are set in the transient analysis setting and the same
seed values are shared as the noise seeds in the transient noise analysis. The
traditional Monte Carlo parameter sampling is used for random values from the
GAUSS, AGAUSS, UNIF, and AUNIF defined parameters. The transient noise
METHOD=MC keyword will inject random noise waveforms for all noise sources
present in the circuit, including SPICE models. The simulation result is an
ensemble of output waveforms that reflect both the traditional Monte Carlo
parameter variations and the random noise waveform injections from transient
noise analysis.

Note: You will receive a warning message about the .TRANNOISE

sweep being ignored.

232 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Controlling Noise

Monte Carlo Output Data

The .TRANNOISE analysis outputs raw data to the *.trpn0 and *.printtr0
files. Formatting of the data output is consistent with that used for *.pn0 and
*.snpn0 data files. These output files organize data according to the sample
number in the Monte Carlo index.
The first sample (index=1) creates a noise-free simulation, that is, disables all
noise sources for this simulation. With index=2, all subsequent simulations use
unique random number seeding to create unique simulation results due to
noise.

Monte Carlo Examples

The following example generates 30 Monte Carlo noise simulations that starts
with a noiseless (index=1) simulation.
.TRANNOISE v(out) SAMPLES=30

The following example generates 20 Monte Carlo noise simulations that starts
with the seed value (that is, index) of 31 for the first simulation.
.TRANNOISE v(out) SAMPLES=20 SEED=31

The following example generates a single noise simulation, with a seed value of
50, and amplifies all noise sources by a factor of 10.
.TRANNOISE v(out) SEED=50 SCALE=10.0

The following example generates six Monte Carlo transient noise simulations
with seed values of 1, 3, 4, 5, 9, and 10. Normalized autocorrelation is
computed for each v(out) output.
.TRANNOISE v(out) SAMPLES=LIST(1,3:5,9:10) AUTOCORRELATION=2

Controlling Noise
Substantial performance improvements can be obtained by removing noise
from the non-critical circuit blocks. It can also be useful to disable various noise
contributors to isolate and identify the dominant circuit noise effects. Transient
noise analysis allows you to ignore noise contributions at the subcircuit level.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 233

N-2017.12
Chapter 11: Transient Noise Analysis
Setting Noise Bandwidth

To ignore the noise contribution, add the NOISE=0 parameter to any subcircuit
instance specified by an X-element. NOISE=0 is treated as a hierarchical
parameter, causing noise to be disabled down through the subcircuit's
hierarchy.
By default, all noise sources are activated at time=0 in transient analysis.
Using the START keyword in the .TRANNOISE command allows the noise
sources to be activated at a specified time.
For example:
.TRAN 10p 10u
.TRANNOISE v(out)
+ fmin=1k fmax=1e8
+ start=2u $ enable noise sources at 2uS

Setting Noise Bandwidth

The FMAX and FMIN keywords in the .TRANNOISE command are used to set
the noise bandwidth of the circuit. You can expect some significant differences
in transient noise analysis performance with FMAX and FMIN changes. Noise
power increases linearly with FMAX and the flicker noise power can scale as 1/
FMIN.
By default, HSPICE estimates the circuit noise bandwidth and uses the
estimated value for FMAX. The estimate is based on the circuit conductance
and capacitance matrices internal to HSPICE.
You can also manually set the FMAX value. Use caution when setting the FMAX
value manually since it has a dramatic effect on transient noise. Since FMAX
controls the amount of energy each noise source can emit, huge values of
FMAX (such as 100G) can result in huge instantaneous noise levels. FMAX is
therefore a critical parameter for controlling the transient noise analysis
performance.
Additionally, because FMAX sets the noise bandwidth of noise sources, it also
controls the maximum time step taken. Therefore, a large FMAX can cause very
slow simulation performance. Setting FMAX should be based on the known
operating and clock frequencies, or technology limits.
The FMIN keyword establishes the lower frequency bound for flicker noise
modeling. You need to use this control only when running very long simulations
to capture the effects of low frequency noise.

234 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
TIE and Spectrum Jitter Measurements from Transient Noise Results

TIE and Spectrum Jitter Measurements from Transient

Noise Results
While transient noise analysis shows the effect of noise on the signal
magnitude, it is also useful to see how the noise affects the timing of the signal.
From the transient noise analysis results, jitter can be measured. The two jitter
measurements are time interval error or TIE, and autocorrelation function. TIE
measures the time-shift behavior relative to a reference signal and is best
measured using WaveView. The autocorrelation function tracks the relative
time-shift behavior of the signal.
The following sections discuss these topics:
■
Measure TIE Jitter and Jitter Spectrum Example
■
Autocorrelation Function Example

Measure TIE Jitter and Jitter Spectrum Example

The following clock buffer example circuit uses a single run Monte Carlo
transient noise analysis. Post-processing is performed using WaveView.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 235

N-2017.12
Chapter 11: Transient Noise Analysis
TIE and Spectrum Jitter Measurements from Transient Noise Results

The complete clockbuf4.sp input file for this example is located at

$installdir/demo/hspice/rf_examples/.
The transient noise analysis uses the following:
■
A .TRAN command and a .TRANNOISE command:
.tran '0.05*period' '520*period'
.trannoise v(outb26) FMAX=50G SCALE=10

The noise bandwidth is set by FMAX and the SCALE factor is used to scale
the noise sources in the circuit.
■
A pulse source to define the input clock source. The pulse source will also
be used as the ideal reference source for comparison with the noisy buffer
output signals.
Vsrc ref gnd DC 0 PULSE (0.0 'vdd'
+ '0.975*period' '0.05*period' '0.05*period'
+ '0.45*period' 'period')
■
A .PROBE command for plotting the output signals:
.probe tran v(ref) v(outa) v(outb14) v(outb26)

Analysis Setup Details

* Transient noise example
.option post probe
.option runlvl=5
.param freq=2000MEG period='1/freq'
* Reference Clock
Vsrc ref gnd DC 0 PULSE (0.0 'vdd'
+ '0.975*period' '0.05*period' '0.05*period'
+ '0.45*period' 'period')
* Clock Buffer Circuit
* Analysis setup
.tran '0.05*period' '520*period'
.trannoise v(outb26) FMAX=50G SCALE=10
.probe tran v(ref) v(outa) v(outb14) v(outb26)
.end

To run this netlist, type the following command:

hspice -i clockbuf4.sp -o

This produces an output file clockbuff4.tr0, containing the transient noise

output.

236 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
TIE and Spectrum Jitter Measurements from Transient Noise Results

To measure the TIE jitter, use the following steps:

1. Type wv at the prompt to invoke Custom WaveView.
2. Use File > Import Waveform File and select the clockbuf4.tr0 file from
the Open: Waveform Files dialog box.
3. Select the v(outb26) signal from the clockbuf4.tr0 file in the signal
browser. Double-click the signal name or drag and drop the signal in the
waveform. The v(outb26) signal is the noisy output of the clock buffer.
4. To plot the jitter versus time, use Tools > Jitter vs. Time. In the Jitter vs.
Time Plot window, set the Reference Clock to An Analog Signal. Set the
External Reference Signal to be the ideal reference by dragging and
dropping the v(ref) signal into the box. Select Display Reference Clock
Signal. Set the Rise (V) and Fall (V) for both the Analog External
Reference Signal Threshold and the Analog Target Signal Threshold to
0.4 volts. The Jitter Type should be set to Trigger Time. Select Always
Search Forward for Target Edge. Set the Jitter Output to Bar Chart. Click
OK. The reference signal, v(ref) and the jitter verses time waveforms will
now be displayed. If you zoom in, the waveforms will look similar to the
following figure:

Figure 35 WaveView: Jitter vs. Time measurement to get Timing Interval Error
(TIE) Jitter vs. Time. Measurement based on the reference signal

HSPICE® User Guide: Advanced Analog Simulation and Analysis 237

N-2017.12
Chapter 11: Transient Noise Analysis
TIE and Spectrum Jitter Measurements from Transient Noise Results

5. To get the jitter spectrum, click on the FFT/DFT icon in the toolbar. In the
FFT/DFT Parameters window, set the # of points to 1024. Deselect
Sampling Rate. The box next to Stop will now be checked. Enter 260n for
Stop. Select Spectrum Panel for Display Type and click OK. The FFT of
the jitter will be displayed in a new WaveView window.
6. Change the Y-axis to dB10 by right clicking the Y-axis and selecting Plot
Complex > dB10(mag) from the menu.
7. Change the spectrum to lines by right clicking inside the WaveView window
and selecting Configure Spectrum from the menu. In the Spectrum
Settings window, select Draw Lines and click OK. The jitter spectrum looks
similar to the following figure:

Figure 36 FFT of the TIE Jitter samples show the Jitter Spectrum

Autocorrelation Function Example

The autocorrelation function is given by the following equation:


R   =  f  t f  t +    dt
–

238 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
TIE and Spectrum Jitter Measurements from Transient Noise Results

The example netlist used to illustrate the measurement of TIE jitter can also be
used to illustrate the use of the autocorrelation function. To activate the
autocorrelation calculation for output terminals, the transient noise analysis
statement is:
.trannoise v(outb26) FMAX=20G SCALE=10 autocorrelation=1

The output data from the autocorrelation function calculation is output to an

ASCII formatted file with the extension .trnz#. The autocorrelation function
can be viewed in WaveView by adding a .PROBE statement to probe ONOISE.
In the example netlist, the probe statement is:
.PROBE trannoise onoise

For the clockbuf4.sp example, the autocorrelation file is named

clockbuf4.trnz0 and it contains the output v(outb26) and the
autocorrelation function of the output, autocorr(outb26). The following figure
shows the output signal and the autocorrelation function plotted in WaveView
and zoomed in to show more detail of the waveforms.

Figure 37 Autocorrelation Measurement

HSPICE® User Guide: Advanced Analog Simulation and Analysis 239

N-2017.12
Chapter 11: Transient Noise Analysis
Phase Noise and Jitter from Transient Noise

Phase Noise and Jitter from Transient Noise

Phase noise and jitter characterization of closed loop PLLs can be difficult
using the HB or SN simulation algorithms. The PLL must be in steady-state and
this requires that the PLL must be simulated well beyond the transient start-up
period. Because of multiple time scales due to dividers that are present in the
typical PLL, the HB and SN algorithms may have convergence issues. With
additional processing, phase noise and jitter measurements can be extracted
from the transient noise simulation.
Extracting the phase noise of a PLL from transient noise analysis can reveal
noise shaping and spurious responses. Two methods are available in transient
noise analysis to characterize the phase noise and jitter. Additionally, the
number of window segments, used by the spectral density estimation algorithm
in transient noise analysis, can be set using the option TRNOISESEGMENTS.
Setting the option results in a smoother frequency domain spectral density
response.
The following sections discuss these topics:
■
Delay Line Method
■
Phase Detector Method
■
Phase Noise Output
■
Jitter Output
■
Recommendations

Delay Line Method

Phase noise can be measured by multiplying a signal by the delayed version of
itself. This results in a spectral density of frequency fluctuations (FM noise) that
is then converted to phase noise.
2 2 2
L  f m  =  S  f  f m     2f m    V rms  f m     2K  2f m  
m

The advantages of this method are less sensitivity to reference frequency

errors and the requirement of a smaller total transient time (TSTOP). The delay
line method is better for measuring the phase noise of VCO circuits.

240 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Phase Noise and Jitter from Transient Noise

Using Delay Line Method

The delay line method is selected by setting phasenoise=1 in the transient
noise command.
A reference source definition is required to establish the phase reference for
phase noise and jitter calculations. HSPICE now provides the capability to
automatically compute the frequency for use as the reference for phase noise
and jitter measurements. The automatic frequency detection algorithm is based
on the time-frequency analysis of signals based on the Ramanujan Fourier
Transform expansion, which has special properties in its ability to estimate
frequency content as a function of time (see [7]).
To activate the automatic frequency detection algorithm, set REF=auto in the
.TRANNOISE command. Alternatively, you can also specify a reference source
present in the netlist or a frequency value (in Hz), for use as the reference for
phase noise and jitter measurements. For detailed information on using REF for
.TRANNOISE, see .TRANNOISE in HSPICE Reference Manual: Commands
and Control Options.
Jitter can be measured and probed by setting jitter=1.
.option RUNLVL=5 $ for phase accuracy
.tran 10p 42u $ TSTOP=42u
Vref refn gnd SIN (0.5 0.5 'VCO_ref') $ reference
Rref refn gnd 1k
.option TRNOISESEGMENTS=4 $ smoothing
.trannoise v(vco_out) FMAX=4G
+ start=2u $ noise begins at 2u
+ ref=Vref $ specify reference source
+ phasenoise=1 $ use delay line method
+ jitter=1 $ extract jitter from phase noise
.probe trannoise phnoise $ probe phase noise
.probe trannoise jitter $ probe TIE jitter

Phase Detector Method

The phase detector method compares the zero crossings between the noisy
signal and an ideal sinusoidal reference signal, making a direct measurement
of phase fluctuations and their spectral density.
1
L  f m  = ---  S   f m  
2

HSPICE® User Guide: Advanced Analog Simulation and Analysis 241

N-2017.12
Chapter 11: Transient Noise Analysis
Phase Noise and Jitter from Transient Noise

This method does not require calibration of the phase detector gain (K) and is
immune to DC offsets, AM noise, and delay values. The phase detector method
is better for measuring the phase noise of PLL circuits.

Using the Phase Detector Method

The delay line method is selected by setting phasenoise=2 in the transient
noise command. The jitter can be extracted from the phase noise by setting
jitter=2.
.option RUNLVL=5 $ for phase accuracy
.tran 10p 42u $ TSTOP=42u
Vref refn gnd SIN (0.5 0.5 'frq_ref*divideN') $ reference
Rref refn gnd 1k
.option TRNOISE_SEGMENTS=4 $ smoothing
.trannoise v(xmc_top.mc_clk_vco_out)
+ samples=1 seed=2 FMAX=4G
+ start=2u $ noise begins at 2u
+ ref=Vref $ specify reference source
+ phasenoise=2 $ use delay line method
+ jitter=2 $ extract jitter from phase noise
.probe trannoise phnoise $ probe phase noise
.probe trannoise jitter $ probe TIE jitter

Phase Noise Output

The keyword phnoise specifies the phase noise for the whole circuit. The
phase noise value for a circuit element can be specified using the
phnoise(element_name) keyword.
.PRINT TRANNOISE phnoise phnoise(element_name)
.PROBE TRANNOISE phnoise phnoise(element_name)

In this syntax, the standalone phnoise is the phase noise parameter.

For example: .PROBE TRANNOISE phnoise.
When phase noise is extracted from transient noise analysis, the phase noise
data is output to the *.trpn# file when .OPTION POST is set and .PROBE
statements are used. The phase noise data is output to the *.printtrpn# file
when .PRINT statements are used. The unit of the phase noise data is
decibels, relative to the carrier signal, per hertz, across the output nodes
specified in the .TRANNOISE statement (Units: dBc/Hz). The phase noise data
plot is a function of the offset frequency.

242 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Phase Noise and Jitter from Transient Noise

Jitter Output
The timing jitter calculations are derived from the results of phase noise
analysis. The phase noise output syntax supports the JITTER keyword as an
output keyword in addition to the PHNOISE keyword.
.PRINT TRANNOISE PHNOISE JITTER
.PROBE TRANNOISE PHNOISE JITTER

When jitter is extracted in addition to phase noise from transient noise analysis
and the JITTER keyword is present in the print and probe statements, the
.TRANNOISE analysis also outputs the raw jitter data to the *.trjt0 file when
.OPTION POST is set and .PROBE statements are used. The raw jitter data is
output to the *.printtrjt0 file when .PRINT statements are used. The jitter
data is the TIE timing jitter (Accumulated Jitter) and is a function of time in units
of seconds.
For a discussion on the different jitter measurements available in HSPICE, see
Jitter Definitions on page 114.

Recommendations
The following recommendations should be followed to get the best phase noise
and jitter results from transient noise analysis.
When simulating a PLL, it is important to know when the PLL has reached the
steady-state. It is recommended that the START parameter in the transient
noise command is used to activate noise calculations, at a time after the
transient analysis has reached steady-state and after the PLL loop dynamics
have stabilized. This will speed up the simulation and prevent errors in
measuring phase noise and jitter caused by a PLL that has not yet stabilized to
its steady-state output frequency.
It is important to know the PLL or VCO output frequency since a SIN or PULSE
source that operates at the PLL or VCO output frequency is used as a
reference source for the phase noise and jitter calculations.
To get the low offset frequency phase noise response, a large TSTOP value in
the transient noise command is required. The recommended TSTOP value is
given as:
TSTOP  Smoothing / lowest_frequency

HSPICE® User Guide: Advanced Analog Simulation and Analysis 243

N-2017.12
Chapter 11: Transient Noise Analysis
Power Spectral Density from Transient Noise

Where the phase noise smoothing is controlled by the option

TRNOISESEGMENTS.
Use caution when using transient noise analysis for open-loop oscillators
because low-Q VCOs may have considerable frequency variations due to
unbounded phase drifts. With unstable frequencies, the phase-detector method
can give erroneous results. Hence, for open-loop oscillators, the delay line
method is a better alternative. For most open-loop VCOs, the best phase noise
analysis is .PHASENOISE based on .SNOSC or .HBOSC analysis.
The .PHASENOISE analysis is designed to be used with autonomous (non-
driven) circuits using .SNOSC and .HBOSC, since it uses frequency domain
methods to achieve faster speed and better dynamic range than time domain
methods. In addition, special windowing functions are used during the spectral
estimation phase to reduce the numerical noise.
Because insufficient accuracy settings may cause numerical noise to look like
phase noise or jitter, consider checking the dynamic range of the noise
measurements. To do this, set SEED=1 in the transient noise command to test
noise measurements with noise disabled. The result will be the numerical noise
floor. To ensure the accuracy of the noise results, look for 10dB noise to noise
ratio. One should always expect a numerical noise floor limit in measurements.

Power Spectral Density from Transient Noise

An important application of transient noise analysis is to generate output
spectra of circuit waveforms that include electronic noise. In cases where
transient noise analysis results are being compared to the HSPICE small signal
noise (.NOISE) analysis results, the desired output measurements are in terms
of the noise power spectral density or PSD. The required outputs are
frequency-domain density responses in the units of V2/Hz and A2/Hz (or V/
sqrt(Hz) and A/sqrt(Hz)) and the total noise effects are computed by integrating
the density values.
To compute PSD values as a function of frequency from time-domain signals,
HSPICE uses special spectral estimation techniques. In these techniques, the
estimation of the noise output level is considered equally important as the
signal level, to extract basic frequency-domain noise power spectral density
output measurement from Monte Carlo transient noise analysis.

244 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Stochastic Differential Equation Analysis

Using PSD from Transient Noise

Power spectral density can be extracted by adding the keyword PSD=1 to the
transient noise analysis command.
An important HSPICE option that can be used to control the PSD calculations
is:
.option TRNOISESEGMENTS=int
Where the value int represents the number of window segments used by
HSPICE for the spectral density estimation algorithm. As this value is
increased, additional Welch noise averaging is used to provide a smoother
frequency domain noise density response. In addition, the .TRANNOISE
START parameter value is used to identify when the circuit has reached steady
state and as the starting time for performing PSD calculations.
.option runlvl=5
.option trnoisesegments=20
.tran 10p 10u
.trannoise v(out) fmax=10g psd=1
.probe trannoise onoise(m)

Stochastic Differential Equation Analysis

Transient noise analysis predicts waveform statistics at particular time points.
One method to accomplish this is to run the multi-sample Monte Carlo
approach and then use the ensemble of simulation results to generate
histograms at the time points of interest. Histograms that measure vertical
distributions reveal voltage noise. Histograms that measure horizontal time-
shift distributions reveal jitter. In those cases where it is desirable to have a
highly accurate distribution curve, it may be necessary to simulate with a large
number of Monte Carlo samples. Stochastic differential equation (SDE)
analysis can provide a more efficient alternative in such cases. Instead of
requiring a large ensemble of results, SDE performs special calculations that
directly predict the statistics of the output waveforms.
Let the time-domain signal resulting from a regular transient analysis (with
s
noise ignored) for a specific output node be written as v out  t  .

Let the time-domain signal resulting from an analysis with signal and noise for
s+n
the same output node be written as v out  t  .

HSPICE® User Guide: Advanced Analog Simulation and Analysis 245

N-2017.12
Chapter 11: Transient Noise Analysis
Stochastic Differential Equation Analysis

s
We can define the noise voltage component v out  t  to be:

Equation 52
n s+n s
v out  t  = v out  t  – v out  t 
We can define a variance equal to the expected value of this noise component
at a given time to be:

Equation 53
2 n n
 n  t  = vout  t   vout  t 
If we assume that the noise variations are small, we can create a linear
Stochastic Differential Equation (SDE) for the noise contribution vector x n t 
n
(one entry in the vector being the output noise v out  t  ). This SDE may be
formulated in terms of time-varying coefficient matrices that are evaluated for a
normal transient analysis, which are functions of the noise-free solution vector
x s  t  and derived from the noise-free transient analysis computations [1]:

Equation 54
A  t x n + C  t x·n + B  t v  0
x n  0  = x n0
We can then create and solve a linear ordinary differential equation (ODE)
T
system for the time-varying noise correlation matrix K  t  = x n  t x n  t  :

Equation 55
· T
K  t  = E  t K  t  + F  t F  t 
where E  t  and F  t  are derived from A  t  , B  t  , and C  t  , and T indicates
the transpose operation.
In general, simultaneously solving for both the deterministic and stochastic
differential equations can give us the complete time-dependent output signal
waveform vector x s  t  , as well as the complete time-varying noise correlation/
T
covariance matrix K  t  = x n  t x n  t  . The entries in this matrix represent time-
2 n
dependent variance values  n  t  for output signals v out  t  . This output can
be interpreted and plotted as a time-varying RMS noise voltage waveforms for

246 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Stochastic Differential Equation Analysis

n 2
v RMS t  =  n  t  . The results of such a transient + noise analysis include the
usual deterministic transient analysis waveforms, including the mean voltage
s
output v out  t  and also its (stochastic) time-varying RMS noise component
n
v RMS t  . In this sense, the SDE analysis method provides typical SPICE
output waveforms for circuit unknowns, plus the additional waveform
representing the time-varying statistics of the circuit. The .TRANNOISE SDE
method therefore activates this special analysis and makes this output
available to the user.
Stochastic Differential Equation (SDE) dynamic noise statistics are computed
in the form of a time-varying covariance matrix. SDE techniques allow the
output of a variance waveform for any output signal. Variance waveforms can
be used to construct probability density plots.
The SDE approach gives probability density information and, like the multi-run
Monte Carlo approach, is useful for characterizing statistics at specific time
events.
The following topics are discussed in these sections:
■
SDE Output Data
■
SDE Examples

SDE Output Data

The .TRANNOISE SDE analysis outputs raw data to the *.tr0 and
*.printtr0 files consistent with transient analysis. The .PRINT/.PROBE
output syntax supports the following measurements:
.print trannoise ONOISE ONOISE(M) VRMS(n1[,n2])
.probe trannoise ONOISE ONOISE(M) VRMS(n1[,n2])
where:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 247

N-2017.12
Chapter 11: Transient Noise Analysis
Stochastic Differential Equation Analysis

■
The ONOISE and ONOISE(M) outputs are the same. They represent noise
voltage or current at the node or branch specified by the output keyword.
The ONOISE represents the RMS noise voltage component (square root of
the variance), units in Volts, of the noise at the specified output present in
addition to the noise-less transient voltage.
Note: ONOISE is only output when you use the SDE method; when
using the Monte Carlo method, ONOISE is 0.
■
VRMS: The output of RMS noise voltages at other nodes (that is, the output
for general nodal noise voltage values).
The actual instantaneous output voltage is the sum of the signal plus noise
components:

Equation 56
s+n s n
v out  t  = v out  t  + v out  t 
n
Where the noise component v out  t  has an assumed Gaussian distribution (in
x ) as:
2
–x
---------2
1 2
-------------- e
 2
s
And the output signal v out  t  is that resulting from the (deterministic) .TRAN
analysis. The time-varying RMS noise voltage waveforms (for onoise) are
n 2
related to the variance at the specified outputs as given by v RMS  t  = n  t 
2 n n
where:  n  t  = v out  t   v out  t  .

SDE Examples
Example 1. SDE method with maximum frequency of 5GHz.
.TRANNOISE v(7) METHOD=SDE FMAX=5g

Example 2. Activates SDE noise analysis and dumps the ONOISE output to the
*.tr0 file:
.TRANNOISE v(out) METHOD SDE
.PROBE TRANNOISE ONOISE

248 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 11: Transient Noise Analysis
Error Handling, Error Recovery, Status Reporting

Example 3. Activates SDE noise analysis placing a lower bound on flicker noise
to be 10kHz and an upper bound on all noise power at 100MHz:
.TRANNOISE v(out) METHOD=SDE FMIN=10k FMAX=100MEG

Error Handling, Error Recovery, Status Reporting

The following error checks are made with transient noise analysis:
■
Verify that the specified output node exists.
■
Verify non-negative integer values for “list” entries.

Note: Invalid values for FMIN and FMAX will be ignored and default
values will be used.

The SDE solving can be substantially slower than transient analysis. Status
reporting includes the progress of the noise analysis.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 249

N-2017.12
Chapter 11: Transient Noise Analysis
References

References
[1] A. Demir, E.W.Y. Liu, A.L. Sangiovanni-Vincentelli, “Time-domain non-
Monte Carlo noise simulation for nonlinear dynamic circuits with arbitrary
excitations,” IEEE Trans. CAD, vol. 15, no. 5. May 1996.
[2] P. Bolcato and R. Poujois, “A new approach for noise simulation in transient
analysis”, Proc. IEEE Int. Symp. Circuits Syst., pp. 887-890, May 1992.
[3] J. A. McNeill, “Jitter in ring oscillators,” Ph.D. dissertation, Boston University,
1994.
[4] M. Okumura and H. Tanimoto, “A time-domain method for numerical noise
analysis of oscillators”, Proc. IEEE Asia Pacific Design Automation
Conference, 1997.
[5] A. van der Ziel, Noise in Solid State Devices and Circuits, John Wiley &
Sons, 1986.
[6] A. Hajimiri, S. Limotyrakis, and T.H. Lee, “Jitter and phase noise in ring
oscillators,” IEEE J. Solid-State Circuits, vol. 34, no. 6, pp. 790-804, June
1999.
[7] P.P. Vaidyanathan, "Ramanujan sums in the context of signal processing",
IEEE Trans. Signal Processing, vol. 62, no. 16, August 15, 2014.

250 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 12
Periodic Time-Dependent Noise Analysis
12

(.PTDNOISE)

Describes periodic time-dependent noise analysis.

This chapter discusses these topics:

■ Introduction to Periodic Time-Dependent Noise Analysis
■
PTDNOISE Input Syntax
■
PTDNOISE Output Syntax and File Format
■ Errors and Warnings
■
Usage Example
■
Troubleshooting .PTDNOISE Simulation
■ References

Introduction to Periodic Time-Dependent Noise

Analysis
While HBNOISE and SNNOISE calculate a time-averaged power spectral
density, there are applications where a characterization of the time-
dependence of the noise is required. These applications include the
computation of the jitter associated with a noisy signal crossing a threshold and
the computation of the noise associated with the discretization of an analog
signal, which computes the noise in a periodically-driven circuit at a point in
time. Periodic Time-Dependent noise analysis (PTDNOISE) calculates the
noise spectrum and the total noise at a point in time. Jitter in a digital threshold
circuit can then be determined from the total noise and the digital signal slew
rate.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 251

N-2017.12
Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
Introduction to Periodic Time-Dependent Noise Analysis

Circuits driven by large periodic signals produce cyclostationary noise, that is,
the noise characteristics are periodic in time. Cyclostationary noise can be
characterized in several ways, with the particular application determining which
is appropriate.[1] The time-average power spectral density (PSD) ignores
frequency correlations in the noise, but is adequate when the fundamental
frequency of the cyclostationary noise is much larger than the bandwidth of
interest. The time-average PSD is calculated in the HBNOISE/SNNOISE
analyses. [2]
The harmonic power spectral density (HPSD) or equivalently, the auto-
correlation function, R(t1,t2), contains the correlation information between
noise sidebands. This information is necessary to build behavioral
cyclostationary noise sources and to separate the amplitude modulation (AM)
and the phase modulation (PM) noise components. (See Amplitude
Modulation/Phase Modulation Separation for more information.)
The time-dependent power spectral density (TDPSD) can be integrated over
frequency to yield the time-dependent noise (TDN). TDN can then be used to
determine the jitter associated with a noisy signal crossing a threshold.
PTDNOISE analysis allows the calculation of TDPSD, TDN, and jitter. In
addition, you can calculate both the time-domain power spectral density
(TDSN) and the integrated noise (time-dependent noise, TDN) at multiple time
points.
By measuring the jitter associated with a noisy signal crossing a threshold, the
jitter is modeled by displacing the time in a noise free signal v(t) with a
stochastic process j(t).

Equation 57
V jitter  t  = v  t + j  t  
We can also determine the voltage at this node including the time-dependent
noise n(t):

Equation 58
Vn  t  = v  t  + n  t 
By equating these two representations, expanding in a Taylor series, and
dropping higher order terms, as follows:

Equation 59
V  t  + n  t  = v  t + j  t   = v  t  + dv  t   dt  j  t  + 

252 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
PTDNOISE Input Syntax

Equation 60
N  t  = dv  t   dt  j  t 
In terms of variances, jitter is then defined as:

Equation 61
2 2
Var  j  t   = n  t    dv  t   dt 

PTDNOISE Input Syntax

.PTDNOISE output TIME=[val|meas|sweep]
+ [TDELTA=time_delta]
+ frequency_sweep
+ [listfreq=(frequencies|none|all)]
+ [listcount=val] [listfloor=val]
+ [listsources=on|off]

Parameter Description

output Is an output node, pair of nodes, or 2-terminal elements. HSPICE advanced analog analyses
references the equivalent noise output to this node (or pair of nodes). Specify a pair of nodes
as V(n+,n-); only one node as V(n+, n-). If you specify only one node, V(n+), then HSPICE
advanced analog analyses assumes the second node is ground. You can also specify a 2-
terminal element name that refers to an existing element in the netlist.

TIME Time point at which time domain noise is evaluated. Specify either a time point explicitly, such
as: TIME=value, where value is either numerical or a parameter name or a .MEASURE name
associated with a time domain .MEASURE command located in the netlist. PTDNOISE uses
the time point generated from the .MEASURE command to evaluate the noise characteristics.
This is useful if you want to evaluate noise or jitter when a signal reaches some threshold value.

TDELTA A time value used to determine the slew rate of the time-domain output signal. Specified as
TDELTA=value. The signal slew rate is then determined by the output signal at TIME +/-
TDELTA and dividing this difference by 2 x TDELTA. This slew rate is then used in the calculation
of the strobed jitter. If this term is omitted a default value of 0.01 x the .SN period is assumed.

frequency_sweep Frequency sweep range for the output noise spectrum. The upper and lower limits also specify
the integral range in calculating the integrated noise value. Specify LIN,DEC, OCT, POI,
SWEEPBLOCK, DATA sweeps. Specify the nsteps, start, and stop frequencies using the
following syntax for each type of sweep:
■ LIN nsteps start stop
■
DECnsteps start stop
■ OCT nsteps start stop
■
POI nsteps freq_values
■ SWEEPBLOCK=swblockname
■
DATA dataname

HSPICE® User Guide: Advanced Analog Simulation and Analysis 253

N-2017.12
Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
PTDNOISE Output Syntax and File Format

Parameter Description

listfreq Prints the element noise value to the .lis file. This information is only printed if a noise
spectrum is requested in a PRINT or PROBE statement. (See PTDNOISE Output Syntax and
File Format.) You can specify which frequencies the element noise is printed. The frequencies
must match the sweep_frequency values defined in the frequency_sweep, otherwise they are
ignored.
In the element noise output, the elements that contribute the largest noise are printed first. The
frequency values can be specified with the NONE or ALL keyword, which either prints no
frequencies or every frequency defined in frequency_sweep. Frequency values must be
enclosed in parentheses. For example:
■
listfreq=(none)
■ listfreq=(all)
■
listfreq=(1.0)
■ listfreq=(1.0G, 2.0G)
The default value is NONE.

listcount Prints the element noise value to the .lis file, which is sorted from the largest to smallest
value. You do not need to print every noise element; instead, you can define listcount to print
the number of element noise frequencies. For example, listcount=5 means that only the top
5 noise contributors are printed. The default value is 1.

listfloor Prints the element noise value to the .lis file and defines a minimum meaningful noise value
(in V/Hz1/2 units). Only those elements with noise values larger than listfloor are printed.
The default value is 1.0e-14 V/Hz1/2.

listsources Prints the element noise value to the .lis file when the element has multiple noise sources,
such as a MOSFET, which contains the thermal, shot, and 1/f noise sources. You can specify
either ON or OFF: ON prints the contribution from each noise source and OFF does not. The
default value is OFF.

PTDNOISE Output Syntax and File Format

PTDNOISE output syntax allows for the output of a single parameter: onoise,
where onoise is the noise voltage spectral density at each frequency point
specified by the frequency_sweep keyword for the time points specified by the
V
TIME keyword. The units are ----------- .
Hz
.PROBE PTDNOISE onoise

254 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
PTDNOISE Output Syntax and File Format

.PRINT PTDNOISE onoise

Parameter Units Description

onoise V- Noise voltage spectral density at each frequency point specified by

---------- frequency_sweep at the time point specified by time_value
Hz

Output File Format

The following PTDNOISE output files are generated depending on the user
input:

File Description

*.printptn# Writes output from the .PRINT statement when using HB to obtain the steady state
solution

*.ptn# Writes output from the .PROBE statement when using HB to obtain the steady state
solution

*.printsnptn# Reports output from the .PRINT statement when using SN to obtain the steady state
solution.

*.snptn# Writes output from the .PROBE statement when using SN to obtain the steady state
solution.

*.lis Standard output file *.lis contains the following information:

■
Performance Statistics Log
■ Number of Nodes
■
Number of FFT Points
■ Number of Equations
■
Memory in use
■
Maximum Krylov iterations
■
Maximum Krylov Dimension
■
Target GMRES Residual
■ Gmres Residual
■
Actual Krylov Iterations taken
■ Frequency (swept input frequency values)

Noise source contributions are listed sequentially and are controlled by the
PTDNOISE command line parameters: listtime, listfreq, listcount, listfloor, and
listsources.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 255

N-2017.12
Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
Errors and Warnings

.MEASURE Syntax and File Format

The syntax for .MEASURE PTDNOISE is:
.MEASURE PTDNOISE result [integnoise|jitter|slewrate]
.MEASURE PTDNOISE allows for the measurement of these parameters:
integnoise, time-point, tdelta-value, slewrate, and strobed jitter.

Parameter Units Description

integnoise V Voltage noise integrated over a frequency range specified by

frequency_range at the time point specified by TIME=val.

slewrate v/sec Output signal slewrate at the time point specified by TIME=val.

jitter sec Calculated from the noise voltage (integrated over the frequency range
specified by frequency_range), divided by the slew rate at the same
node(s), at the time point specified by TIME=val.

Note: .MEASURE PTDNOISE is ignored when a TIME=sweep is

specified in the netlist and a warning message is issued.

Measure File Format

File Description

*.msnptn# Contains output from the .MEASURE statement when using .SN to obtain the steady
state solution.

Errors and Warnings

Error messages are generated under the following circumstances:
■
PTDNOISE frequency sweep includes negative frequencies. PTDNOISE
allows only frequencies that are greater than or equal to zero.
■
PTDNOISE time sweep includes negative times. PTDNOISE allows only
time points that are greater than or equal to zero.
■ No SN statement is specified (error at parser). PTDNOISE requires an SN
statement to generate the steady-state solution.
■
Incorrect match to .MEASURE statement.

256 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
Usage Example

A warning is issued for a PTDNOISE convergence failure. When the GMRES

solver reaches the maximum number of iterations and the residual is greater
than the specified tolerance, PTDNOISE generates a warning and then
continues as if the data were valid. The warning reports the following
information:
■
Final GMRES residual
■
Target GMRES residual
■
Maximum Krylov iterations
■
Actual Krylov iterations taken

Usage Example
The following test case illustrates the PTDNOISE analysis for a simple inverter.
* Simple RC + Inverter - rcInvPTDNoise.sp
* rrd Jan 03, 2007
* Simulates PSD(t,f) of a simple inverter
* sweep time points
.param f0 = 5.0e8
.sn tones=f0 nharms=4 trinit=10n
.PTDNOISE v(out1) TIME=lin 3 0 2n TDELTA=.1n dec 5 1e5 1e10
+ listfreq=(1e6,1e8)
+ listcount=1
+ listsources=ON

.MEASURE PTDNOISE strobejit STROBEJITTER onoise FROM = 1e4 TO =

1e10
$.measure SN t1 trig AT=0 targ v(out1) val=1.5 fall=1

.opt post
.probe ptdnoise onoise
.print ptdnoise onoise
.probe sn v(out1)

vd vdd 0 3.0
.global vdd
vgate in0 0 COS(1.5 1.4 'f0' 0 0 0)
rin in0 in1 50
rout out1 0 .1g

xo1 in1 out1 inv

HSPICE® User Guide: Advanced Analog Simulation and Analysis 257

N-2017.12
Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
Troubleshooting .PTDNOISE Simulation

.subckt inv in out

m1 out in 0 0 n l=350e-9 w=4.5e-6
m2 out in vdd vdd p l=350e-9 w=4.5e-6
.ends

.MODEL N NMOS
+Level= 49 Tnom=27.0 version =3.1 TLEVC= 1
*
***

.MODEL P PMOS
+Level= 49 Tnom=27.0 version =3.1 TLEVC= 1

.end

Troubleshooting .PTDNOISE Simulation

When performing jitter measurements on a clock buffer using PTDNOISE
analysis, you may get unusual results. If you define noise frequencies that are
multiples of your input clock, you get large values in your jitter measurement.
When you sweep through frequencies that are multiples of the clock, PTD noise
up-converts frequency=0 flicker noise. (These contributions can be large,
–n
since flicker noise typically has a freq dependence where n  1 .) For this
reason you should avoid sweeping through multiples of the clock frequency.
Also, the flicker noise is limited at very small frequencies so you do not see
infinite noise.

258 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
References

References
[1] J. Roychowdhury, D. Long, and P. Feldmann, “Cyclostationary Noise
Analysis of Large RF Circuits with Multitone Excitations,” IEEE Journal of
Solid-State Circuits, volume 33, pages 324–336, March 1998.
[2] A. Demir, A. Sangiovanni-Vincentelli, “Analysis and Simulation of Noise in
Nonlinear Electronic Circuits and Systems”, Kluwer Academic, 1998.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 259

N-2017.12
Chapter 12: Periodic Time-Dependent Noise Analysis (.PTDNOISE)
References

260 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Part 4: Behavioral Modeling
Applications

This Part contains the following chapters/topics.

■ Chapter 13, Behavioral Modeling of Nonlinear Circuits
■
Chapter 14, Laplace Transform Modeling of Filters and Networks

HSPICE® User Guide: Advanced Analog Simulation and Analysis 261

N-2017.12
262 HSPICE® User Guide: Advanced Analog Simulation and Analysis
N-2017.12
 13
Behavioral Modeling of Nonlinear Circuits
13

Describes how to create and use behavioral models.

Behavioral modeling substitutes more abstract, less computationally intensive,

circuit models for lower-level descriptions of analog functions. These simpler
models emulate the transfer characteristics of the circuit elements that they
replace, but with increased efficiency. Behavioral modeling substantially
reduces the actual simulation time per circuit. At the level of an entire design
and simulation cycle, design efficiency greatly increases, and you can complete
a design (from conception to a marketable product) in substantially less time.
HSPICE ships numerous examples for your use; see Behavioral Application
Examples for paths to demo files.
This chapter presents the following topics:
■
Behavioral Design Process
■
Using Behavioral Elements
■ Voltage and Current Controlled Elements
■
Modeling with Digital Behavioral Components
■
Calibrating Digital Behavioral Components
■
Analog Behavioral Elements
■
Op-Amps, Comparators, and Oscillators
■
Phase-Locked Loops (PLL)
■ References

HSPICE® User Guide: Advanced Analog Simulation and Analysis 263

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Behavioral Design Process

Behavioral Design Process

HSPICE provides specific modeling elements that promote the use of
behavioral and mixed signal techniques. These models include controllable
sources that you can configure, to emulate op-amps, single-input or multi-input
logic gates, or any system with a continuous algebraic transfer function.
■
You can create these functions in algebraic form, or in the form of coordinate
pairs.
■ You can use digital stimulus files to enter logic waveforms into the simulation
deck, rather than using piecewise linear sources to enter digital waveforms.
■
You can define the clock rise times, fall times, periods, and voltage levels.
With HSPICE behavioral models, the typical design cycle for a circuit or system
is:
1. Fully simulate a subcircuit, with pertinent inputs, characterizing its transfer
functions.
2. Determine which HSPICE elements, singularly or in combination,
accurately describe the transfer function.
3. Reconfigure the subcircuit appropriately.
4. After you verify the behavioral model, substitute the model into the larger
system, in place of the lower-level subcircuit.

Using Behavioral Elements

Behavioral elements offer a higher level of abstraction and faster processing,
compared to a lower-level description of an analog function.
■
System-level designers can use function libraries of subcircuits containing
these elements, to describe parts such as:
• Op-amps
• Vendor-specific output buffer drivers
• TTL drivers
• Logic-to-analog converters

264 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Using Behavioral Elements

• Analog-to-logic simulator converters

■
Integrated Circuit designers can use these elements to reduce the design
time, especially when designing filters and signal processors.
Behavioral elements use an arbitrary algebraic equation as a transfer function
for either a voltage (E) or a current (G) source. This function can include:
■
Nodal voltages
■
Element currents
■
Time
■
Other parameters which you define
A good example of this is a VCO, where control is the input voltage node,
and osc is the oscillator output:
Evco osc 0 VOL=’voff+gain*\\
SIN(6.28*freq*(1+V(control))*TIME)’

You can use subcircuits to encapsulate a function.

■
If you split the function definition from the use, you create a hierarchy.
■
If you pass parameters into the subcircuit, you create a parameterized cell.
■
If you create a full transistor cell library and a behavioral representation
library, you can include mixed-signal functions within HSPICE.
You can use the built-in OPTIMIZE function to calibrate the behavioral
elements from a full transistor circuit.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 265

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Using Behavioral Elements

Accuracy High Level Speed

Full Optimization
Transistor Behavioral
Subcircuit Subcircuit

Signal Mode

Circuit Representation

Figure 38 Netlisting by Signal Mode

The following sections discuss these topics:

■
Controlled Sources
■
Libraries

Controlled Sources
Controlled sources model both analog and digital circuits, at the behavioral
level. This reduces simulation times for mixed signals and models system-level
operations. Controlled sources also model gate-switching action for the
behavioral modeling of digital circuits. For analog behavioral modeling, you can
program the controlled sources as mathematical functions. These functions
can be either linear or non-linear, depending on other nodal voltages and
branch currents.

Libraries
The Discrete Device Library contains standard industry IC components. You
can use this library to model board-level designs that contain any of the
following:
■ Transistors
■
Diodes

266 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Voltage and Current Controlled Elements

■
Opamps
■
Comparators
■
Converters
■
IC pins
■
Printed circuit board traces
■
Coaxial cables
You can also model drivers and receivers, to analyze transmission line effects,
power line noise, and signal line noise.

Voltage and Current Controlled Elements

HSPICE provides two voltage-controlled and two current-controlled elements,
known as E, F, G, and H-elements. For a description of these elements, see
Sources and Stimuli, in HSPICE® User Guide: Elements.

Modeling with Digital Behavioral Components

The following sections show how to model using digital behavioral components:
■
Behavioral AND and NAND Gates
■
Behavioral D-Latch
■
Behavioral Double-Edge Triggered Flip-Flop

Behavioral AND and NAND Gates

The following example uses a G Element to model a 2-input AND gate. An E-
element models a two-input NAND gate. Figure 39 shows the resulting
waveforms. This example is located in the following directory:
$installdir/demo/hspice/behave/behave.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 267

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Modeling with Digital Behavioral Components

Figure 39 NAND/AND Gates

Behavioral D-Latch
This example uses one input NAND gate and NPWL/PPWL functions, to model a
D flip-flop.

268 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Modeling with Digital Behavioral Components

QB

CLK

A Q

DATA CLKN

CLKN

RX

CLK

D – LATCH

Figure 40 D-Latch

Example
This example is located in the following directory:
$installdir/demo/hspice/behave/dlatch.sp
The file contains the following examples:
■
Waveforms
■
Subcircuit definitions for Behavioral N-Channel MOSFET
■
Behavioral P-Channel MOSFET

HSPICE® User Guide: Advanced Analog Simulation and Analysis 269

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Modeling with Digital Behavioral Components

Figure 41 D-Latch Response

Behavioral Double-Edge Triggered Flip-Flop

This example uses the D_LATCH subcircuit from the previous example and
several NAND gates, to model a double-edged, triggered flip-flop.
Example
This example is located in the following directory:
$installdir/demo/hspice/behave/det_dff.sp
■
Main Circuit
■
Subcircuit Definitions

270 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Modeling with Digital Behavioral Components

D-Latch
Clck
Clock R1
Q W2 Q
D Clock
D Q

W1
D-Latch QB
Clock R2
Clckn Q W3
Clock
D Q
Q3

Figure 42 Double-Edge Triggered Flip-Flop

Figure 43 Double Edge Triggered Flip-Flop Response

HSPICE® User Guide: Advanced Analog Simulation and Analysis 271

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

Calibrating Digital Behavioral Components

The following sections discuss these topics:
■
Building Behavioral Lookup Tables
■
Optimizing Behavioral CMOS Inverters
■
Optimizing Behavioral Ring Oscillators

Building Behavioral Lookup Tables

The following simulation demonstrates an ACL family output buffer, with 2ns
delay, and 1.8ns rise and fall time. It also shows ground and VDD supply
currents, and internal ground bounce due to the package.

vdd

OUT
D

Figure 44 ACL Family Output Buffer

The following commands automatically measure the datasheet quantities, such

as TPHL, risetime, maximum power dissipation, and ground bounce.

272 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

.MEAS tphl trig v(D) val=’.5*vdd’ rise=1

+ targ v(out) val=’.5*vdd’ fall=1
.MEAS risetime trig v(out) val=’.1*vdd’ rise=1
+ targ v(out) val=’.9*vdd’ rise=1
.MEAS max_power max power
.MEAS bounce max v(xin.v_local)

The inverter consists of capacitors, diodes, one-dimensional lookup table

MOSFETs, and a special low-pass delay element. A property of the low-pass
delay element is to attenuate pulses that are narrower than the delay value.

V+

IN OUT
Delay

V-

Figure 45 Inverter

Subcircuit Definition
.subckt inv in out v+ v-
cout+ out_l v+ 2p
cout- out_l v- 2p
xmp out_l inx v+ pmos
xmn out_l inx v- nmos
e inx v- delay in v- td=1n
din v- in dx
.model dx d cjo=2pf
chi in v+ .5pf
.ends inv

HSPICE® User Guide: Advanced Analog Simulation and Analysis 273

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

One-dimensional lookup tables represent the behavioral MOSFETs.

Behavioral N-Channel MOSFET

The following example is a Drain Gate source.
.subckt nmos 1 2 3
gn 3 1 VCR npwl(1) 2 3 scale=0.008
* VOLTAGE RESISTANCE
+ 0. 495.8840g
+ 200.00000m 456.0938g
+ 400.00000m 141.6902g
+ 600.00000m 7.0624g
+ 800.00000m 258.9313meg
+ 1.00000 6.4866meg
+ 1.20000 842.9467k
+ 1.40000 21.6882k
+ 1.60000 170.8367k
+ 1.80000 106.4944k
+ 2.00000 72.7598k
+ 2.20000 52.4632k
+ 2.40000 38.5634k
+ 2.60000 8.8056k
+ 2.80000 5.2543k
+ 3.00000 4.3553k
+ 3.40000 3.4950k
+ 3.80000 2.0534k
+ 4.20000 2.7852k
+ 4.60000 2.5k
+ 5.0 2.3k
.ends nmos

The preceding example is a voltage-versus-resistance table. It shows, for

example, that the resistance at 5 V is 2.3 kohms.

Creating a Behavioral Inverter Lookup Table

You can create an inverter lookup table in three simple steps:
1. Simulate an actual transistor level inverter, using a DC sweep of the input.
2. Print the V/I output, for the output pullup and pulldown transistors.
3. Copy the printed output into the volt lookup table element, for the controlled
resistor.
The following test file, inv_vin_vout.sp, calculates RN (the effective
pulldown resistor transfer function) and RP (the pullup transfer function).

274 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

■
RN is calculated as Vout/I(mn), where mn is the pulldown transistor.
■
RP is calculated as (VCC-Vout)/I(mp), where mp is the pullup transfer
function.
The actual calculation uses a more accurate method to obtain the series
resistance of the transistor, as in Figure 46.

Vdx Rtot= (Vds-Vsx)/Ids

RD For greater accuracy:

Vd Rtot= RD + RS + (vd-vs)/Ids
Vs RD = 1/LV16(mn)

RS RS = 1/LV17(mn)
(vd-vs) = LX3(mn)
Vsx Ids = LX4(mn)

Figure 46 VIN versus VOUT

The first graph in Figure 47 shows VIN versus VOUT.

The second graph shows the computed transfer resistances (RP and RN), as a
function of VIN.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 275

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

5.0 INV.VCB.TOB
OUT
4.0
Volt [Lin]

3.0

2.0

1.0

9.9e-1
9.0e-1 INV.VCB.TOB
1.0e-1 RB
1.0e-1
Volt [Lin]

RB
110.4
10.4
1.4
110.4X
10.0X
1.4X
110.4K
10.0K
1.0K
500.0 0 1.0 2.0 3.0 4.0 5.0
Volts [Lin]

Figure 47 RP and RN as a Function of VIN

The HSPICE file used to calculate RP and RN is located in the following

directory:
$installdir/demo/hspice/behave/inv_vin_vout.sp

Optimizing Behavioral CMOS Inverters

To calibrate behavioral models, run HSPICE on the full transistor version of a
cell. Then optimize the behavioral model to this data.

276 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

VCC

in out

in out
Gd
Cin Rin Et 1K Rout Cout

Figure 48 CMOS Inverter and its Equivalent Circuit

In this example, HSPICE uses the LEVEL 3 MOSFET model to simulate the
CMOS inverter.
1. To obtain the input and output resistances, HSPICE performs a .TF transfer
function analysis (.TF V(out) Vin).
2. To obtain the transfer function table of the inverter, HSPICE performs the DC
analysis and sweeps the input voltage (.DC Vin 0 5 .1).
3. HSPICE uses this table, in the PWL element, to represent the transfer
function of the inverter.
4. A voltage-controlled PWL capacitance adjusts the rise and fall time of the
inverter, in the equivalent circuit, across the output resistance.
5. The delay element obtains the propagation delay across the output RC
circuit.
6. HSPICE uses the inverter in a ring oscillator, to adjust the input capacitance.
7. HSPICE uses optimization analysis for all adjustments in this example. The
data file and the results are shown.
Example
This example is located in the following directory:
$installdir/demo/hspice/behave/invb_op.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 277

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Calibrating Digital Behavioral Components

The invb_op.sp file contains the following sections:

■
Subcircuit Definition
■
Inverter Using Model
■
Optimization Results
■
Optimization Completed
■
Optimized Parameters OPTINV
■
Optimize Results Measure Names and Values

Figure 49 CMOS Inverter Response

Optimizing Behavioral Ring Oscillators

To optimize the behavioral ring oscillator performance, review the examples in
this section.
Example: Five-Stage Ring Oscillator
This example is located in the following directory:
$installdir/demo/hspice/behave/ring5bm.sp
The ring5bm.sp file also contains the results of the five-stage ring oscillator
example.

278 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Analog Behavioral Elements

Figure 50 Ring Oscillator Response

Analog Behavioral Elements

The following components are examples of analog behavioral building blocks.
Each component demonstrates a basic HSPICE feature:
■
integrator: ideal op-amp E-element source
■
differentiator: ideal op-amp E-element source
■ ideal transformer: ideal transformer E-element source
■ AM modulator: algebraic G-element source
■
data sampler: algebraic E-element source
HSPICE uses an ideal op-amp to model the integrator circuit and a VCVS to
adjust output voltage. The following equation calculates output of the integrator:
gain
Vout = – ----------------   Vin  dt + Vout  0 
t
Ri  Ci 0

HSPICE® User Guide: Advanced Analog Simulation and Analysis 279

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Analog Behavioral Elements

Ri Ci

Vin Vout
Egain
+ - +
- out1 -
•
+

Figure 51 Integrator

Figure 52 Response of Integrator to a Triangle Waveform

Example
This example is located in the following directory:
$installdir/demo/hspice/behave/integ.sp
The integ.sp file also contains the following sections:

280 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Analog Behavioral Elements

■
Control and options
■
Subcircuit definition
■
Circuit
The following sections discuss these topics:
■
Behavioral Differentiator
■
Behavioral Amplitude Modulator
■
Behavioral Data Sampler

Behavioral Differentiator
HSPICE uses an ideal op-amp to model a differentiator and a VCVS to adjust
the magnitude and polarity of the output. The following equation calculates the
differentiator response:

Vout = – gain  Rd  Cd  d Vin

dt
For a high-frequency signal, the output of a differentiator can overshoot the
edges. To smooth this out, you can use a simple RC filter.

Cd
Rd R

Vin Vout
Egain
+ - + C
- out1 -
•
+

Figure 53 Differentiator

Example
This example is located in the following directory:
$installdir/demo/hspice/behave/diff.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 281

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Analog Behavioral Elements

The diff.sp file also contains the following sections:

■
Control and Options
■
Subcircuit Definition
■
Circuit

Figure 54 Response of a Differentiator to a Triangle Waveform

Behavioral Amplitude Modulator

This example, which uses a G-element as an amplitude modulator with a pulse
waveform carrier, is located in the following directory:
$installdir/demo/hspice/behave/amp_mod.sp
See also AM Modulation in HSPICE User Guide: Basic Simulation and
Analysis.

282 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Analog Behavioral Elements

Figure 55 Amplitude Modulator Waveforms

Behavioral Data Sampler

An example of sampling the behavioral data is located in the following
directory:
$installdir/demo/hspice/behave/sampling.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 283

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

Figure 56 Sampled Data

Op-Amps, Comparators, and Oscillators

The following sections discuss these topics:
■
741 Op-Amp from Controlled Sources
■
Inverting Comparator with Hysteresis
■ Voltage-Controlled Oscillator (VCO)
■
LC Oscillator

741 Op-Amp from Controlled Sources

To model the A741 operational amplifier, use PWL controlled sources. A
piecewise linear CCVS (source “h”) limits the output to ±15 V.

284 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

rin+ out1 r1 r2 r3 r4 out2 r0 out

in+
g c r e c1 c2 c3 c4 eo h
rin + + +
- - -

in-
rin- I(g) = F(Vin+ - Vin-)
e = V(out1)
eo = V(out2)
V(out) = F ( I(h) )

Figure 57 Operational Amplifier

Example
This example is located in the following directory:
$installdir/demo/hspice/behave/op_amp.sp
The op_amp.sp file also contains the following sections:
■
Main Circuit
■
RC Circuit With Pole At 9 MHz
■
Output Limiter to 15 V

HSPICE® User Guide: Advanced Analog Simulation and Analysis 285

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

Figure 58 AC Analysis Response

Figure 59 Transient Analysis Response

286 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

Inverting Comparator with Hysteresis

A piecewise-linear VCVS models an inverting comparator.

Vout
Vout
a
Vohigh
Vin +- + Ecomp
-
-2u 1u

b Volow Vab

Rf
Rb Cb
Open loop characteristic of
comparator Ecomp

Figure 60 Inverting Comparator with Hysteresis

Two reference voltages correspond to the volow and vohigh voltages of Ecomp:
Volow  Rb Vohigh  Rb
Vreflow = ---------------------------- Vrefhigh = ------------------------------
Rb + Rf Rb + Rf
When Vin exceeds Vrefhigh, the Vout output changes to Volow. For Vin values
less than Vreflow, the output changes to Vohigh.
An example is located in the following directory:
$installdir/demo/hspice/behave/compar.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 287

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

1.0 COHPAE.TRO
Volt [Lin] VIN

0 VOUT

-1.0
0 500.0 1.00 1.90 2.00
Time [Lin]
-2.0 COHPAE.TRO
Volt [Lin]

VOUT

0 -3.0 -2.0 -1.0 0 1.0 2.0 3.0

VCN [Lin]
5.0
COHPAE.TRO
Volt [Lin]

VOUT

0 -3.0 -2.0 -1.0 0 1.0 2.0 3.0

VOB [Lin]

Figure 61 Response of Comparator

Voltage-Controlled Oscillator (VCO)

In this example, a one-input NAND (functioning as an inverter) models a five-
stage ring oscillator. PWL capacitance switches the load capacitance of this
inverter from 1pF to 3pF. As the simulation results indicate, the oscillation
frequency decreases as the load capacitance increases.
Example
This example is based on the demonstration netlist vcob.sp, which is
available in the $installdir/demo/hspice/behave directory. This file
also contains a sample subcircuit definition.

288 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

vcob.sp voltage controlled oscillator using pwl functions

.option post
.global ctrl
.tran 1n 100n
.ic v(in)=0 v(out1)=5
.probe tran v(in)
x1 in out1 inv
x2 out1 out2 inv
x3 out2 out3 inv
x4 out3 out4 inv
x5 out4 in inv
vctrl ctrl 0 pwl(0,0 35n,0 40n,5)
*
* macro definitions
*
.subckt inv in out rout=1k
gcout out 0 pwl(1) ctrl 0 level=2 delta=.01
+ 4.5 1p
+ 4.6 3p
rout out 0 rout
gn 0 out nand(1) in 0 scale='1.0k/rout'
+ 0. 5.00ma
+ 0.25 4.95ma
+ 0.5 4.85ma
+ 1.0 4.75ma
+ 1.5 4.42ma
+ 3.5 1.00ma
+ 4.000 0.50ma
+ 4.5 0.20ma
+ 5.0 0.05ma
.ends inv
*

.end

HSPICE® User Guide: Advanced Analog Simulation and Analysis 289

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

Figure 62 Voltage Controlled Oscillator Response

LC Oscillator
The initial capacitor charge is 5 V. The value of the capacitance is the function
of voltage at node 10. The capacitance value becomes four times higher at the
t2 time. The following equation calculates the frequency of this LC circuit:
1
freq = -------------------------------
6.28  L  C
At the t2 time, the frequency must be halved. The amplitude of oscillation
depends on the condition of the circuit when the capacitance value changes.
The stored energy is:
2 2
E =  0.5  C  V  +  0.5  L  I 
2 2
E = 0.5  C  Vm  I = 0 E = 0.5  L  Im  V = 0
At the t2 time, when V=0, if C changes to A  C, then:
2 2 2
0.5  L  Im = 0.5  Vm = 0.5   A  C   Vm

290 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

and from the above equation:

Vm
Vm = --------
A

Qm = A  Vm
The second condition that HSPICE considers is when V=Vin, if C changes to A
C, then:
Qm = Qm
C  Vm = A  C  Vm
Vm
Vm = --------
A
Therefore, HSPICE modifies the voltage amplitude between Vm/sqrt(A) and
Vm/A, depending on the circuit condition when the circuit switches. This
example tests the CTYPE=0 and 1 results. The result for CTYPE=1 must be
correct because capacitance is a function of voltage at node 10, not a function
of the voltage across the capacitor itself.
Example
This example is based on demonstration netlist calg2.sp which is available in
the $installdir/demo/hspice/behave directory.
* in this example the ctype 0 and 1 is tested. the result for
* ctype=1 must be correct because capacitance is function of
* voltage at node 10, not voltage across itself.
*
.option post
.ic v(1)=5 v(2)=5
c1 1 0 c='1e-9*v(10)' ctype=1
l1 1 0 1m
*
c2 2 0 c='1e-9*v(10)' ctype=0
l2 2 0 1m
*
v10 10 0 pwl(0sec,1v t1,1v t2,4v)
r10 10 0 1

HSPICE® User Guide: Advanced Analog Simulation and Analysis 291

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Op-Amps, Comparators, and Oscillators

COLOR.TOP
4.0 VCL
Volt [Lin]
2.0

-2.0

-4.0

10.0 COLOR.TOP
Q1
5.0
Volt [Lin]

-5.0

-10.0 0 10.0 20.0 30.0 40.0 50.0 60.0

Time [Lin]

Figure 63 Correct Result Corresponding to CTYPE=1

COLOR.TOP
4.0
VC2
Volt [Lin]

2.0

-2.0

-4.0

40.0
COLOR.TOP
Q2
30.0
Volt [Lin]

20.0

10.0

-10.0
0 10.0 20.0 30.0 40.0 50.0 60.0
Time [Lin]

Figure 64 Incorrect Result Corresponding to CTYPE=0

292 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Phase-Locked Loops (PLL)

Phase-Locked Loops (PLL)

The following sections discuss phase-locked loops:
■
Phase Detector, with Multi-Input NAND Gates
■
Phase Locked Loop Modeling

Phase Detector, with Multi-Input NAND Gates

This circuit uses behavioral elements to implement the inverters, with 2, 3, and
4 input NAND gates.

HR
R U1

Q1

H4
Q2

Q3

Q4
Y D1

NY
V2
YN
D2

Figure 65 Phase Detector

Example
An example is located in the following directory:
$installdir/demo/hspice/behave/pdb.sp

HSPICE® User Guide: Advanced Analog Simulation and Analysis 293

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Phase-Locked Loops (PLL)

This file also contains sample subcircuit definitions.

Figure 66 Phase Detector Response

Phase Locked Loop Modeling

A phase-locked loop (PLL) circuit synchronizes to an input waveform, within a
selected frequency range. This returns an output voltage that is proportional to
the variations in the input frequency. It has three basic components:
■
A voltage-controlled oscillator (VCO) which returns an output waveform that
is proportional to its input voltage.
■ A phase detector which compares the VCO output to the input waveform
and returns an output voltage depending on their phase difference.
■
A loop filter which filters the phase detector voltage. Returns an output
voltage which forms the VCO input (and external voltage output) of the PLL.

294 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Phase-Locked Loops (PLL)

Phase Detector Loop Filter

InB In Out Rloop Cloop

OutB Rloop

Signal Out In

Signal OutB InB

Voltage Controlled Oscillator

Figure 67 Behavioral Phase-Locked Loop

The PLL can be implemented using behavioral elements (Figure 67) or using
bipolar transistors (Figure 68 on page 296 and Figure 69 on page 297).
The netlist for the behavioral PLL is in the pll_bvp.sp file and the netlist for
the full bipolar PLL is in the pll.sp file. The netlist for the full bipolar PLL
contains the loop filter and the output circuit. Both these netlist files are
available in the directory: $installdir/demo/hspice/behave.
The PLL transfer function shows a linear region of voltage vs. (periodic) time
which is defined as the “lock” range.
The results of transient simulations (Figure 68) show minimal difference
between implementations. However, run time statistics show that the behavioral
model reduces the simulation time, to one-third that of the full circuit.
If you use this PLL in a larger system simulation (for example, an AM tracking
system), include the behavioral model. This model substantially reduces the
simulation run time and still accurately represents the subcircuit.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 295

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
Phase-Locked Loops (PLL)

Figure 68 Behavioral (PLL_BVP Curve) vs. Bipolar (PLL_Curve) Simulation

296 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 13: Behavioral Modeling of Nonlinear Circuits
Phase-Locked Loops (PLL)

VCC
1K

625
Out

In 300

In2 VEE
.5mA

300
In2B
OutB

625
InB

1K

Figure 69 Bipolar Phase Detector

HSPICE® User Guide: Advanced Analog Simulation and Analysis 297

N-2017.12
Chapter 13: Behavioral Modeling of Nonlinear Circuits
References

References
[1] Chua & Lin. Computer Aided Analysis of Electronic Circuits. Englewood
Cliffs: Prentice-Hall, 1975, page 117. See also “SPICE2 Application Notes
for Dependent Sources,” by Bert Epler, IEEE Circuits & Devices Magazine,
September 1987.

298 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 14
Laplace Transform Modeling of Filters and
14

Networks

Describes modeling filters and networks, including Laplace transforms.

When you apply Kirchhoff’s laws to circuits that contain energy storage
elements, the result is simultaneous differential equations in the time domain. A
simulator must solve these equations, to analyze the circuit’s behavior. Solving
any equation that is higher than the first order can be difficult and classical
methods cannot easily solve some driving functions.
In both cases, to simplify the solution, you can use Laplace transforms. These
transforms convert time domain equations, containing integral and differential
terms, into algebraic equations in the frequency domain.
HSPICE ships numerous examples for your use; see Filters Examples in
HSPICE User Guide: Basic Simulation and Analysis for paths to the demo files.
The following sections discuss these topics:
■
Transient Modeling
■
Using G- and E-elements
■ Laplace and Pole-Zero Modeling
■
Modeling Switched Capacitor Filters
■
References

Transient Modeling
The Laplace transform method provides an easy way to relate a circuit’s
behavior in time and frequency-domains. This facilitates simultaneous work in
those domains.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 299

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Transient Modeling

The algorithm that the HSPICE simulator uses for Laplace and pole/zero
transient modeling, offers better performance than the Fast Fourier Transform
(FFT) algorithm. To invoke Laplace and pole/zero transient modeling, use a
LAPLACE or POLE function call in a source element statement.
Laplace transfer functions are especially useful in top-down system design,
when you use ideal transfer functions instead of detailed circuit designs. In
HSPICE, you can also mix Laplace transfer functions with transistors and
passive components. Using this capability, you can model a system as the sum
of the contributing ideal transfer functions. You can then progressively replace
these functions with detailed circuit models, as they become available.
Conventional uses of Laplace transfer functions include control systems and
behavioral models that contain non-linear elements.
Laplace transforms reduce the time needed to design and simulate large
interconnect systems such as clock distribution networks. You can use the
asymptotic waveform evaluation (AWE) and other methods, to create a Laplace
transfer function model. The AWE model can use only a few poles to represent
the large circuit. You can input these poles through a Laplace transform model,
to closely approximate the delay and overshoot characteristics of many
networks, within a fraction of the original simulation time.
You can use pole/zero analysis to help determine the stability of the design.
You can use the POLE function in HSPICE when the poles and zeros of the
circuit are specified, or you can use the .PZ statement (see .PZ in HSPICE
Reference Manual: Commands and Control Options) to derive the poles and
zeros from the transfer function.
Frequency response is an important analog circuit property. It is normally the
ratio of two complex polynomials (functions of complex frequencies), with
positive real coefficients. The form of frequency response can be either the
locations of poles and zeros, or a frequency table.
The usual way to design complex circuits is to interconnect smaller functional
blocks of known frequency responses, either in the pole/zero or the frequency
table form. For example, to design a band-reject filter, you can interconnect a
low-pass filter, a high-pass filter, and an adder. Study the function of the
complex circuit in terms of its component blocks, before you design the actual
circuit. After you test the functionality of the component blocks, you can use
these blocks as a reference in optimization techniques, to determine the value
of the complex element.

300 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

Using G- and E-elements

The following sections discuss these topics:
■
Laplace Transform Function Call
■
G- and E-element Notes
■
Laplace Band-Reject Filter
■
Laplace Low-Pass Filter
■
Circular Convolution Example

Laplace Transform Function Call

Use the G- and E-elements as linear functional blocks or as elements with
specific frequency responses.
In the following equations, H(s) denotes the frequency response (also called
the impulse response), where s is a complex frequency variable ( s = j2f ). To
obtain the frequency response, perform an AC analysis and set AC=1 in the
input source (the Laplace transform of an impulse is 1). The following
expression relates the input and output of the G- and E-elements, with the
specified frequency response:
Y  j2f  = Hj  2f   Xj  2f 
where X is the input, Y is the output, and H is the transfer function, at the f
frequency.
AC analysis uses the above relation, at any frequency, to determine the
frequency response. For operating point and DC sweep analysis, the relation is
the same but the frequency is zero.
The transient analysis is more complicated than the frequency response. The
output is a convolution of the input waveform with the impulse response h(t):
t

y t =  x  ht –   d
–

In discrete form, the output is:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 301

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

y  k  =   x  m   h   k – m     , k = 0, 1, 2, ...
m=0

where you can obtain h(t) from H(f), using the inverse Fourier integral:


h t =  H  f   e j2ft  df
–

The following equation calculates the inverse discrete Fourier transform:

N–1 j2nm
1 ----------------
h  m  = ------------
N  H  fn   e N , m = 0, 1, 2, ..., N-1
n=0

where N is the number of equally-spaced time points and  is the time interval
or time resolution.
For the frequency response table form (FREQ) of the LAPLACE function,
HSPICE uses a performance-enhanced algorithm, to convert H(f) to h(t). This
algorithm requires N to be a power of 2. The following equation determines the
fn frequency point:

n
f n = ------------ , n = 0, 1, 2, ..., N-1
N
where n > N/2 represents the negative frequencies. The following equation
determines the Nyquist critical frequency:
1
f c = f N/2 = -----------
2
Because the negative frequency responses are the image of the positive
responses, you need to specify only N/2 frequency points, to evaluate N time
points of h(t). The larger the value of fc, the more accurate the transient
analysis results. However, for large fc values, the  becomes smaller and the
computation time increases.
The maximum frequency of interest depends on the functionality of the linear
network. For example, in a low-pass filter, you can set fc to the frequency at
which the response drops by 60 dB (a factor of 1000).
H max
H  f c  = ---------------
1000

302 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

After you select or calculate fc, the following equation can determine :

1
 = -----------
2  fc
The following equation calculates the frequency resolution:
1
f = f 1 = ------------
N
which is inversely proportional to the maximum time (N.), over which HSPICE
evaluates h(t). Therefore, the transient analysis accuracy also depends on the
frequency resolution, or the number of points (N).
You can specify the frequency resolution (DELF) and the maximum frequency
(MAXF) in the G- or E-element statement. To calculate N, HSPICE uses 2
MAXF/DELF. Next, HSPICE modifies N as a power of 2. The effective DELF is 2
MAXF/N, to reflect the changes in N.

G- and E-element Notes

■
If elements in the data file specify frequency responses, do not use pole/
zero analysis. If you specify MAXF=<par> in a G- or E-element Statement,
HSPICE warns that it is ignoring MAXF. This is normal.
■
HSPICE performs circular convolution, when the G-element ACCURACY =
0 or 1 (see Figure 76 on page 308).

Laplace Band-Reject Filter

This example models an active band-reject filter with 3-dB points at 100 and
400 Hz and <35 dB of attenuation between 175 and 225 Hz. The band-reject
filter contains low-pass and high-pass filters, and an adder. The low-pass and
high-pass filters are fifth-order Chebyshev, with 0.5-dB ripple.

Low-Pass

Input Output
S

High-Pass

Figure 70 Band-Reject Filter

HSPICE® User Guide: Advanced Analog Simulation and Analysis 303

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

Example
This example is located in the following directory:
$installdir/demo/hspice/filters/BandstopL.sp
The BandstopL.sp file also contains a sample band-reject filter circuit.

1.0H BOWSTOP.0
VPIOUT
800.0H

600.0H

400.0H

200.0H
Volt [Lin]

0
180.0 BOWSTOP.0
VPIOUT
100.0

-100.0

-180.0 10.0 100.0 3.0 ??0.0

Hertz [Lin]

Figure 71 Frequency Response of the Band-Reject Filter

1.0 BOWSTOP.0
VCIN
800.0H
VCOUT
600.0H

400.0H

200.0H
Volt [Lin]

-200.0H

-400.0H

-600.0H

-800.0H

-1.0
0 2.0H 4.0H 6.0H 8.0H 10.0H 12.0H
Time [Lin]

Figure 72 Transient Response of the Band-Reject Filter to a 250 Hz Sine Wave

304 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

Laplace Low-Pass Filter

This example simulates a third-order low-pass filter with a Butterworth transfer
function. It also compares the results of both the actual circuit and the
functional G Element, with the third-order Butterworth transfer function, for AC
and transient analysis.

3.546F -
1 1
+ Out
In 1
1.392F 0.2024F

Figure 73 Third-Order Active Low-Pass Filter

The third-order Butterworth transfer function that describes the above circuit is:
1.0
H  s  = ----------------------------------------------------------------------------------------------------------------------------------------------------
1.0   s + 1   s + 0.5 + j2  0.1379   s + 0.5 –  j2  0.1379  
The following example is the input listing for the above filter. Parameters set the
pole locations for the G-element. Also, this listing specifies only one of the
complex poles. The program derives the conjugate pole. The output of the
circuit is the out node and the output of the functional element is outg.
Example
An example of a third-order low-pass Butterworth filter is located in the
following directory:
$installdir/demo/hspice/filters/Low_Pass.sp
The Low_Pass.sp file also contains a sample circuit description.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 305

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

0 LOW_POSE.TOB
PDBOUT
-20.0
PDBOUT
-40.0

-60.0

-80.0
Volt [Lin]

-100.0

-120.0 LOW_POSE.TOB
PDBOUT
-130.0
PDBOUT

-100.0

-180.0 14.0h 100.0H 1.0H 10.0H

Time [Lin]

Figure 74 Frequency Response of Circuit and Functional Element

1.20 LOW_POSE.TOB
VCOUT

1.0 VCOUT

800.0H
Volt [Lin]

600.0H

400.0H

200.0H

-200.0H
0 50.0 100.0 150.0 200.0
Time [Lin]

Figure 75 Transient Response of Circuit and Functional Element to a Pulse

306 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

Circular Convolution Example

This 30-degree phase-shift filter uses circular convolution. If DELF=10 MHz,
HSPICE uses inverse fast Fourier transform (IFFT) to obtain the period of time
domain response for the G-element. This value is based on the input frequency
table and is 100 ns. The FREQ G-element performs the convolution integral
from t - T to t, assuming that all control voltages at t<0 are zero. t is the target
time point and T is the period of the time domain response, for the G-element.
In this example, during time points from 0 to 100 ns, HSPICE uses harmonic
components higher than 10 MHz due to the input transition at t=0. So the circuit
does not behave as a phase shift filter. After one period (t>100 ns), HSPICE
performs circular convolution based on a period of 100 ns. The transient result
represents a 30-degree phase shift, for continuous periodic control voltage.

Notes
■
V(ctrl): control voltage input.
■
V(expected): node. Represents an ideal 30-degree shifted wave for the
input.
■
V(test): output of the G Element.

30-Degree Phase Shift Circuit File

This example illustrates a 30-degree phase-shift filter. It is based on the
demonstration netlist phaseshift.sp which is available in the directory:
$installdir/demo/hspice/filters.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 307

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Using G- and E-elements

****
.tran 0.1n 300n
.OPTION post ingold=2 accurate
Vctrl ctrl gnd sin (0 1 10e6)
Gtest gnd test freq ctrl gnd
+ 1.0e00 0 30
+ 1.0e01 0 30
+ 1.0e02 0 30
+ 1.0e03 0 30
+ 1.0e04 0 30
+ 1.0e05 0 30
+ 1.0e06 0 30
+ 1.0e07 0 30
+ 1.0e08 0 30
+ 1.0e09 0 30
+ 1.0e10 0 30
+ MAXF=1.0e9 DELF=10e6
Rtest test gnd 1
Iexpected gnd 3 sin (0 1 10e6 0 0 30)
Vmes 3 expected 0v
Rexpected expected gnd 1
.end

1.50

1.00

0.50 Test

0 30.tro
0 ctrl
expected
-0.50

-1.00

-1.50
0 60.00 0.12 0.18 0.24 0.30
Time [S]

Figure 76 Transient Response of the 30 Degree Phase Shift Filter

308 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Laplace and Pole-Zero Modeling

The following sections discuss these topics:
■
Laplace Transform (LAPLACE) Function
■
Laplace Transform POLE (Pole/Zero) Function
■
AWE Transfer Function Modeling
■
Y-parameter Line Modeling
■
Comparison of Circuit and Pole/Zero Models

Laplace Transform (LAPLACE) Function

HSPICE provides two types of LAPLACE function calls: one for
transconductance and one for voltage gain transfer functions. See Using G-
and E-elements on page 301 for the general forms. For descriptions of the
parameters, see G-element Parameters and E-element Parameters in the
HSPICE® User Guide: Elements.
The following sections discuss these topics:
■
General Form of the Transfer Function
■
Finding the Transfer Function
■
Determining the Laplace Coefficients

General Form of the Transfer Function

To use the LAPLACE modeling function, you must find the k0, ..., kn and d0, ...,
dm coefficients of the transfer function. The transfer function is the s-domain
(frequency domain) ratio of the output for a single-source circuit to the input,
with initial conditions set to zero. The following equation represents the Laplace
transfer function:
Ys
H  s  = -----------
Xs
where:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 309

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

■
s is the complex frequency, j2f.
■
Y(s) is the Laplace transform of the output signal.
■
X(s) is the Laplace transform of the input signal.
Note: To obtain the impulse response H(s), HSPICE performs an
AC analysis, where AC=1 represents the input source. The
Laplace transform of an impulse is 1. For an element with an
infinite response at DC (such as a unit step function H(s)=1/
s), HSPICE calculations use the value of the EPSMIN option
(the smallest number possible on the platform) as the transfer
function.

The general form of the transfer function H(s) in the frequency domain is:

k0 + k1 s +  + kn sn
Hs = --------------------------------------------------
-
d0 + d1 s +  + dm sm
The order of the numerator for the transfer function cannot be greater than the
order of the denominator. The exception is differentiators, for which the transfer
function H(s) = ks. You can use parameter values for all k and d coefficients of
the transfer function, in the circuit descriptions.

Finding the Transfer Function

The first step in determining the transfer function of a circuit is to convert the
circuit to the s-domain. To do this, transform the value for each element into its
s-domain equivalent form.
Table 3 on page 310 and Table 4 on page 312 show transforms that convert
some common functions to the s-domain. The next section provides examples
of using transforms to determine transfer functions.
Table 3 Laplace Transforms for Common Source Functions

f(t), t>0 Source Type L{f(t)}= F(s)

(t) impulse 1

u(t) step 1
---
s

310 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Table 3 Laplace Transforms for Common Source Functions (Continued)

f(t), t>0 Source Type L{f(t)}= F(s)

t ramp 1
----
2
s

e-at exponential 1-
----------
s+a

sin t sine w -
----------------
2 2
s +w

cos t cosine s
-----------------
2 2
s +w

sin(t+ )
sine s sin    +  cos   -
---------------------------------------------
2 2
s +

cos(t+ ) cosine s cos    –  sin   -

---------------------------------------------
2 2
s +

sinht hyperbolic sine  -

----------------
2 2
s –

cosht hyperbolic cosine s -

----------------
2 2
s –

te-at damped ramp 1 -

------------------
2
s + a

e-at sint damped sine 

--------------------------------
2 2
s + a + 

e-at cost damped cosine s+a -

-------------------------------
2 2
s + a + 

HSPICE® User Guide: Advanced Analog Simulation and Analysis 311

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Table 4 Laplace Transforms for Common Operations

f(t) L{f(t)} = F(s)

Kf  t  KF  s 

f1  t  + f2  t  – f3  t  +  F1  s  + F2  s  – F3  s  + 

d sF  s  – f  0– 
ft
dt

2
s F  s  – sf  0  – d f 0– 
2
d ft
2 dt
dt

n n n–1 n–2d
d ft s Fs – s f0 – s f0
n dt
dt

t –1
Fs f 0
– f  t  dt ----------- + --------------
s s

f  t – a u  t – a  a  0 e
– as
Fs
(u is the step function)

e
– at
ft Fs + a

f  at  a  0 1  s
--- F ---
a  a

tf  t  d
– (F  s )
ds

n n
t ft – 1  d F  s 
n
n
ds

312 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Table 4 Laplace Transforms for Common Operations (Continued)

f(t) L{f(t)} = F(s)

 t -
f------- 
t s F  u  du (u is the step function)
f  t – t1  e
–t1 s
Fs

Determining the Laplace Coefficients

The following examples describe how to determine the appropriate coefficients
for the Laplace modeling function call:
■ Laplace Example 1 – Voltage Gain Transfer Function
■
LAPLACE Example 2 – Differentiator
■
LAPLACE Example 3 – Integrator

Laplace Example 1 – Voltage Gain Transfer Function

To find the voltage-gain transfer function for the circuit in Figure 77, convert the
circuit to its equivalent s-domain circuit and solve for vo / vg.

n1 R1= 1000 n2
i1

R2 = 250
vg = 2 sin 3t i2 i3 C = 1F vo
L = 50 mH

n0

Figure 77 LAPLACE Example 1 Circuit

Use transforms fromTable 4 on page 312 to convert the inductor, capacitor, and
resistors. L{f(t)} represents the Laplace transform of f(t):

HSPICE® User Guide: Advanced Analog Simulation and Analysis 313

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

 d 
L  L f  t   = L   sF  s  – f  0   = 50  10   s – 0  = 0.05s
–3
dt
 

1 t  1 Fs f 0
–1 6
L  ----  f  t  dt  = ----   ----------- + -------------- = ----------
1 1
--- + 0 = --------
10

C 0  C  s s  10
–6  s  s

L  R1  f  t   = R1  F  s  = R1 = 1000 W
L  R2  f  t   = R2  F  s  = R2 = 250 W
To convert the voltage source to the s-domain, use the sin  t transform from
Table 3 on page 310:
3 - 6
L  2 sin 3 t  = 2  ---------------
2 2
= -------------
2
-
s +3 s +9
Figure 78 displays the s-domain equivalent circuit.

n1 1000  n2
i1

250 
6 6
v g = ---------------- 10 vo
2 i2 i3 ----------
s +9 s
0.05s

n0

Figure 78 S-Domain Equivalent of the LAPLACE Example 1 Circuit

Summing the output currents from the n2 node:

vo – vg vo vo s
---------------- + ---------------------------
- + -------
- = 0
1000 250 + 0.05s 10 6

Solve for vo:

314 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

1000  s + 5000 v g
v o = ----------------------------------------------------
2
-
6
s + 6000s + 25  10
The voltage-gain transfer function is:
6
vo 1000  s + 5000  5  10 + 1000s -
H  s  = ----- = ----------------------------------------------------
2
- = ----------------------------------------------------
6 6 2
vg s + 6000s + 25  10 25  10 + 6000s + s
For the Laplace function call, use the kn and dm coefficients for the transfer
function, in the form:
n
k0 + k1 s +  + kn s
H  s  = --------------------------------------------------
m
-
d0 + d1 s +  + dm s
The coefficients from the above voltage-gain transfer function are:
k0 = 5 x 106 k1 = 1000

d0 = 25 x 106d1 = 6000 d2 = 1
Using these coefficients, the following is a Laplace modeling function call for
the voltage-gain transfer function of the circuit in Figure 77 on page 313:

LAPLACE Example 2 – Differentiator

To model a differentiator, use either G or E elements, as shown in the following
example.
In the frequency domain:
E-element: V out = ksV in

G-element: I out = ksV in

In the time domain:

dV in
E-element: v out = k
dt

dV in
G-element: i out = k
dt
For a differentiator, the voltage gain transfer function is:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 315

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

V out
H  s  = ---------- = ks
V in
In the general form of the transfer function:

k0 + k1 s +  + kn sn
H  s  = --------------------------------------------------m-
d0 + d1 s +  + dm s

If you set k1 = k and d0 = 1 and the remaining coefficients are zero, then the
equation becomes:
ks
H  s  = ----- = ks
1
Using the k1 = k and d0 = 1 coefficients in the Laplace modeling, the circuit
descriptions for the differentiator are:
Edif out GND LAPLACE in GND 0 k / 1
Gdif out GND LAPLACE in GND 0 k / 1

LAPLACE Example 3 – Integrator

You can use G or E elements to model an integrator, as follows:
In the frequency domain:
k
E Element: V out = -- V in
s
k
G Element: I out = -- V in
s
In the time domain:

E Element: v out = k  V in dt

G Element: i out = k  V in dt

For an integrator, the voltage gain transfer function is:

V out k
H  s  = ---------- = --
V in s
In the general form of the transfer function:

316 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

n
k0 + k1 s +  + kn s
H  s  = --------------------------------------------------m-
d0 + d1 s +  + dm s
As in the previous example, if you set k0 = k and d1 = 1, then the equation
becomes:
k+0++0 k
H  s  = --------------------------------- = --
0+s++0 s

Laplace Transform POLE (Pole/Zero) Function

The following sections describe the general form of the pole/zero transfer
function. It also provides examples of converting specific transfer functions into
pole/zero circuit descriptions.
The following sections discuss these topics:
■
POLE Function Call
■
General Form of the Transfer Function
■ Reduced Form of the Transfer Function
■
RC Line Modeling

POLE Function Call

You can use the POLE function if the poles and zeros of the circuit are available.
You can derive the poles and zeros from the transfer function as described in
this chapter, or you can use the .PZ statement to find them, as described in
.PZ in the HSPICE Reference Manual: Commands and Control Options.
HSPICE provides two forms of the LAPLACE function call: one for
transconductance and one for voltage gain transfer functions. See Using G-
and E-elements on page 301 for the general forms and for optional parameters.
To use the POLE pole/zero modeling function, find the a, b, f, and  coefficients
of the transfer function. The transfer function is the s-domain (frequency
domain) ratio of the output, for a single-source circuit, to the input, with initial
conditions set to zero.

General Form of the Transfer Function

The general expanded form of the pole/zero transfer function H(s) is:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 317

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

a  s +  z1 + j2f z1   s +  z1 – j2f z1   s +  zn + j2f zn   s +  zn – j2f zn 

H  s  = -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
b  s +  p1 + j2f p1   s +  p1 – j2f p1   s +  pm + j2f pm   s +  pm – j2f pm 
You can use parameters to set the a, b, , and f values.
The following is an example:
Ghigh_pass 0 out POLE in 0 1.0 0.0,0.0 / 1.0 0.001,0.0
Elow_pass out 0 POLE in 0 1.0 / 1.0, 1.0,0.0 0.5,0.1379

The Ghigh_pass statement describes a high-pass filter with the transfer

function:
1.0   s + 0.0 + j  0.0 
H  s  = -----------------------------------------------------------
1.0   s + 0.001 + j  0.0 
The Elow_pass statement describes a low-pass filter with the transfer
function:
1.0
H  s  = ----------------------------------------------------------------------------------------------------------------------------------------------------
1.0   s + 1   s + 0.5 + j2  0.1379   s + 0.5 –  j2  0.1379  
To write a pole/zero circuit description for an element, you need to know H(s)
transfer function of the element in terms of the a, b, f, and  coefficients.
Before you use the values of these coefficients in POLE function calls (in the
circuit description), you must simplify the transfer function as described in the
next section.

Reduced Form of the Transfer Function

Complex poles and zeros occur in conjugate pairs (a set of complex numbers
differ only in the signs of their imaginary parts):
 s +  pm + j2f pm   s +  pm – j2f pm  , for poles.

 s +  zn + j2f zn   s +  zn – j2f zn  , for zeros.

To write the transfer function in pole/zero format, supply coefficients for one
term of each conjugate pair. HSPICE provides the coefficients for the other
term. If you omit the negative complex roots, the result is the reduced form of
the transfer function, Reduced{H(s)}.
To find the reduced form, collect all general-form terms that have negative
complex roots:

318 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

a  s +  z1 + j2f z1   s +  zn + j2f zn  a  s +  z1 – j2f z1   s +  zn – j2f zn 

H  s  = ----------------------------------------------------------------------------------------------------  ----------------------------------------------------------------------------------------------------
b  s +  p1 + j2f p1   s +  pm + j2f pm  b  s +  p1 – j2f p1   s +  pm – j2f pm 
Then discard the right-hand term, which contains all terms with negative roots.
What remains is the reduced form:
a  s +  z1 + j2f z1   s +  zn + j2f zn 
Reduced  H  s   = ----------------------------------------------------------------------------------------------------
b  s +  p1 + j2f p1   s +  pm + j2f pm 
For this function, find the a, b, f, and  coefficients to use in a POLE function,
for a voltage-gain transfer function. The following examples show how to
determine the coefficients and write POLE function calls for a high-pass filter
and a low-pass filter.
See the following examples:
■
POLE Example 1 – High-Pass Filter
■
POLE Example 2 – Low-Pass Filter

POLE Example 1 – High-Pass Filter

For a high-pass filter with a transconductance transfer function, such as:
s
H  s  = ---------------------------
 s + 0.001 
Find the a, b,  , and f coefficients needed to write the transfer function in the
general form shown previously. You can then see the conjugate pairs of
complex roots. You need to supply only one of each conjugate pair of roots in
the Laplace function call. HSPICE automatically inserts the other root.
To transform the function into a form that is more similar to the general form of
the transfer function, rewrite the transconductance transfer function as:
1.0  s + 0.0 
H  s  = ----------------------------------
1.0  s + 0.001 
Because this function has no negative imaginary parts, it is already in the
HSPICE reduced form (reference number 2) shown previously.
You can now identify the a, b, f, and  coefficients, so that the H(s) transfer
function matches the reduced form. This matching process obtains the values:
n = 1, m = 1
a = 1.0 z1 = 0.0 fz1 = 0.0

HSPICE® User Guide: Advanced Analog Simulation and Analysis 319

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

b = 1.0 p1 = 0.001 fp1 = 0.0

Using these coefficients in the reduced form, provides the transfer function:
s
--------------------------
-
 s + 0.001 
So the general transconductance transfer function POLE function call:
Gxxx n+ n- POLE in+ in- a  z1,fz1...  zn,fzn /
b  p1,fp1...  pm,fpm

for an element named Ghigh_pass, becomes:

Ghigh_pass gnd out POLE in gnd 1.0 0.0,0.0 /
+ 1.0 0.001,0.0

POLE Example 2 – Low-Pass Filter

For a low-pass filter with the following voltage-gain transfer function:
1.0
H  s  = --------------------------------------------------------------------------------------------------------------------------------------------------------------
1.0  s + 1.0 + j2  0.0   s + 0.5 + j2  0.15   s + 0.5 – j2  0.15 
you need to find the a, b, , and f coefficients, to write the transfer function in
the general form, so that you can identify the complex roots with negative
imaginary parts.
To separate the reduced form, Reduced{H(s)}, from the terms with negative
imaginary parts, rewrite the voltage-gain transfer function as:
1.0 1.0
H  s  = -----------------------------------------------------------------------------------------------------------  --------------------------------------------------
1.0  s + 1.0 + j2  0.0   s + 0.5 + j2  0.15   s + 0.5 – j2  0.15 
1.0
= Reduced  H  s    --------------------------------------------------
 s + 0.5 – j2  0.15 
So:
1.0
Reduced  H  s   = -------------------------------------------------------------------------------------
1.0  s + 1.0   s + 0.5 + j2   0.15 
or:
a  s +  z1 + j2f z1   s +  zn + j2f zn  1.0
---------------------------------------------------------------------------------------------------- = -----------------------------------------------------------------------------------------------------------
b  s +  p1 + j2f p1   s +  pm + j2f pm  1.0  s + 1.0 + j2  0.0   s + 0.5 + j2  0.15 
Now assign coefficients in the reduced form, to match the specified voltage
transfer function. The following coefficient values produce the transfer function:

320 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

n = 0, m = 2,
a =1.0 b = 1.0 p1 = 1.0 fp1 = 0 p2 = 0.5 fp2 = 0.15

You can substitute these coefficients in the POLE function-call, for a voltage-
gain transfer function:
Exxx n+ n- POLE in+ in- a z1,fz1...zn,fzn /
b p1,fp1...pm,fpm

for an element named Elow_pass, to obtain the following statement:

Elow_pass out GND POLE in 1.0 / 1.0 1.0,0.0 0.5,0.15

RC Line Modeling
Most RC lines can use very simple models, with only a single dominant pole.
You can use AWE methods to find the dominant pole, computed based on the
total series resistance and capacitance or determined using the Elmore delay.
The Elmore delay uses the (d1-k1) value as the time constant, for a single-pole
approximation to the complete H(s), where H(s) is the transfer function of the
RC network for a specified output. The inverse Laplace transform of h(t) is H(s):

 DE = 0 t  h  t  dt
Actually, the Elmore delay is the first moment of the impulse response and so
corresponds to a first-order AWE result.

(1) 200 (2) 80 (3) 160 (4) 200 (5)

(6)

v1 +
- 0.6pF 0.8pF 0.7pF 0.8pF
+
e1 -

Figure 79 Circuits for a RC Line

HSPICE® User Guide: Advanced Analog Simulation and Analysis 321

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Example
This example is based on demonstration netlist rcline.sp which is available
in the directory: $installdir/demo/hspice/filters.
****
.Tran 0.02ns 3ns
.OPTION Post Accurate List Probe
v1 1 0 PWL 0ns 0 0.1ns 0 0.3ns 5 1.3ns 5 1.5ns 0
r1 1 2 200
c1 2 0 0.6pF
r2 2 3 80
c2 3 0 0.8pF
r3 3 4 160
c3 4 0 0.7pF
r4 4 5 200
c4 5 0 0.8pF
e1 6 0 LAPLACE 1 0 1 / 1 1.16n
.Probe v(1) v(5) v(6)
.Print v(1) v(5) v(6)
.End

To closely approximate the output of the RC circuit (shown in Figure 79 on

page 321), you can use a single-pole response, as shown in Figure 80 on
page 323.

322 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

5.0 PC.TRO
v(1
4.750
input, v(1) v(5
4.50
v(6
4.250
4.0
3.750
3.50
3.250
full RC line
3.0
2.750
Volt [Lin]

2.50 single-pole
approximation
2.250
2.0
1.750
1.50
1.250
1.0
750.0m
500.0m
250.0m
0
0 500.0p 1.0n 1.50n 2.0n 2.5n 3.0n
Time [Lin]

Figure 80 Transient Response of the RC Line and Single-Pole Approximation

In Figure 80, the single-pole approximation has less delay: 1 ns, compared to
1.1 ns for the full RC line model, at 2.5 V. The single-pole approximation also
has a lower peak value than the RC line model. All other things being equal, a
circuit with a shorter time constant results in less filtering and allows a higher
maximum voltage value. The single-pole approximation produces a lower
amplitude (and less delay) than the RC line because the single pole neglects
the other three poles in the actual circuit. However, a single-pole approximation
still provides very good results for many problems.

AWE Transfer Function Modeling

Approximations using single-pole transfer functions can cause larger errors for
low-loss lines than for RC lines because lower resistance allows ringing. Circuit
ringing creates complex pole pairs in the transfer function approximation. You
need at least one complex pole pair to represent low-loss line response.
Figure 81 on page 324 is a typical low-loss line and the transfer function

HSPICE® User Guide: Advanced Analog Simulation and Analysis 323

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

sources used to test various approximations. To obtain the transfer functions,

HSPICE uses asymptotic waveform evaluation.

(1) 25 (2) 10nH (3) 10nH (4) 100nH (5)

+ 400
v1 1pF 1pF 1pF
-

(6) (7) (8)

+ + +
e1 e2 e3
- - -

Figure 81 Circuits for a Low-Loss Line

The sample file located in the following directory is a low-loss line circuit file:
$installdir/demo/hspice/filters/lowloss.sp
Figure 82 on page 325 shows the transient response of a low-loss line. It also
shows E-element Laplace models using one, two, and four poles. The single-
pole model shows none of the ringing of the higher-order models. Also, all E
models must adjust the gain of their response for the finite load resistance, so
the models are not independent of the load impedance. The 0.94-gain
multiplier in the models take care of the 25-ohm source and the 400-ohm load-
voltage divider. These approximations are good delay estimations.

324 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

full model, v(5) four-pole

Ic2.tr0
6.0 two-pole v(8
model, input, v(1)
v(7
5.50
v(6
5.0 v(5
v(1
4.50

4.0
single-pole
model, v(8)
3.50
Volt [Lin]

3.0

2.50

2.0

1.50

1.0

500.0m

0
0 1.0n 2.0n 3.0n 4.0n 5.0n 6.0n 7.0n 8.0n
Time [Lin]

Figure 82 Transient Response of the Low-Loss Line

Although the two-pole approximation provides a reasonable agreement with the

transient overshoot, the four-pole model offers almost perfect agreement. The
actual circuit has six poles. You can use scaling to bring some of the very small
numbers in the Laplace model above the 1e-28 limit of HSPICE. The SCALE
parameter multiplies every Y-parameter in the LAPLACE specification by the
same value (in this case 1.0e-20).
A low-loss line allows reflections between the load and source, compared to the
loss of an RC line, which usually isolates the source from the load. So you can
either incorporate the load into the AWE transfer function approximation, or
create a HSPICE device model that allows source/load interaction. If you allow
source/load interaction, you do not need to perform the AWE expansions each
time you change load impedances. This allows HSPICE to handle non-linear
loads and removes the need for a gain multiplier, as in the circuit file shown.
You can use four voltage-controlled current sources or G Elements, to create a
Y-parameter model for a transmission line. The
Y-parameter network provides the needed source/load interaction. The next
example shows such a Y-parameter model for a low-loss line.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 325

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

Y-parameter Line Modeling

A model that is independent of load impedance is more complicated. You can
still use AWE techniques, but you need a way for the load voltage and current
to interact with the source impedance. For a transmission line of 100 ohms and
0.4 ns total delay (as shown in Figure 83), to compare the response of the line,
use a Y-parameter model and a single-pole model.

Zk = 100
delay = 4.0 ns/meter
length = 0.1
(1) 25 (2) (5)

400
25 (3) (8)
Y model (6)

+ 400
v1 +
- e1
-

Figure 83 Line and Y-parameter Modeling

Figure 84 shows the voltage and current definitions for a Y-parameter model.

I1 I2
+ Y11 Y12 +

Vin Vout
- Y21 Y22 -

Figure 84 Y Matrix for the Two-Port Network

The following equations describe the general network in Figure 84, which you
can translate into G-elements:
I 1 = Y 11 V in + Y 12 V out

326 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

I 2 = Y 21 V in + Y 22 V out
Figure 85 shows a schematic for a set of two-port Y-parameters. The circuit
consists mostly of G-elements.

Vin I1 I2 Vout

Gy11 Gy12 Gy21 Gy22

I1 = y11Vin + y12Vout
I2 = y21Vin + y22Vout GND

Figure 85 Schematic for the Y-parameter Network

A Pade expansion of the Y-parameters for a transmission line determines the

Laplace parameters for the Y-parameter model, as shown in matrix form in the
following equation:

Y = -----  coth  p  – csch  p 

1
Z o – csch  p  coth  p 

where p is the product of the propagation constant times the line length.
A Pade approximation contains polynomials in both the numerator and the
denominator. A Pade approximation can model both poles and zeros, and coth
and csch functions also contain both poles and zeros. So, a Pade
approximation provides a better low-order model than a series approximation
does.
The following equation calculates the Pade expansion of coth(p) and csch(p),
with a second-order numerator and a third-order denominator:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 327

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

 1 + 2---  p 2  1 – -----1- 2
p
 5   20 
coth  p   ------------------------------- csch  p   -------------------------------
 p + ----- 1- 3
p   p + ----- 7- 3
p 
 15  60

When you substitute ( s  length  LC ) for p, HSPICE generates polynomial

expressions for each G-element. When you substitute 400 nH for L, 40 pF for
C, 0.1 meter for length, and 100 for Zo, ( Z o = L  C ) in the matrix equation
above, you can use the resulting values in a circuit file.
The circuit file shown in the following sections uses all of the equation
substitutions. The Pade approximations have different denominators for csch
and coth but the circuit file contains identical denominators. Although the actual
denominators for csch and coth are only slightly different, using them can
cause oscillations in the HSPICE response. To avoid this problem, use the
same denominator in the coth and csch functions in the example. The
simulation results might vary, depending on which denominator you use as the
common denominator because the coefficient of the third-order term changes
(but by less than a factor of 2).
This sample LC Line circuit file is located in the following directory:
$installdir/demo/hspice/filters/lcline.sp
Figure 86 on page 329 compares the output of the Y-parameter model with that
of a full transmission line simulation, and with that obtained for a single-pole
transfer function. In the latter case, the gain for the load impedance is incorrect,
so the function produces an incorrect final voltage level. As expected, the
Y parameter model provides the correct final voltage level. Although the
Y parameter model provides a good approximation of the circuit delay, it
contains too few poles to model all of the transient details. However, the Y-
parameter model provides excellent agreement with the overshoot and settling
times.

328 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

6.9602 Icpade.tr0
Y-parameter v(8
approximation, v(8)
6.50 v(6
6.0 U-element v(5
transmission line, v(1
5.50

5.0

4.50

4.0
3.50
Volt [Lin]

single-pole
approximation, v(6)
3.0

2.50
input, v(1)
2.0

1.50

1.0

500.0m

0 1.0n 2.0n 3.0n 4.0n 5.0n

Time [Lin]

Figure 86 Transient Response of the Y-parameter Line Model

Comparison of Circuit and Pole/Zero Models

This example simulates a ninth-order low-pass filter circuit and compares the
results with its equivalent pole/zero description using an E-element. The results
are identical, but the pole/zero model runs about 40% faster.The example
shown in Simulation Time Summary on page 329 shows the total CPU times
for the two methods. For larger circuits, the computation time saving can be
much higher.
Figure 87 on page 330 and Figure 88 on page 331 display the transient and
frequency response comparisons resulting from the two modeling methods.

Simulation Time Summary

Circuit model simulation times:

HSPICE® User Guide: Advanced Analog Simulation and Analysis 329

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Laplace and Pole-Zero Modeling

analysis time # points # iter conv.iter

op point 0.23 1 3
ac analysis 0.47 151 151
transient 0.75 201 226 113 rev=0
readin 0.22
errchk 0.13
setup 0.10
output 0.00
total cpu time 1.98 seconds

Pole/zero model simulation times:

analysis time # points #
iter conv.iter
op point 0.12 1 3
ac analysis 0.22 151 151
transient 0.40 201 222 111 rev=0
readin 0.23
errchk 0.13
setup 0.02
output 0.00
total cpu time 1.23 seconds

1owpass9a.1
800.0m v(out
600.0m
400.0m
Volt [Lin]

200.0m
0
-200.0m
-400.0m
-600.0m
-800.0m
ninth.tr0
800.0m v(outp

600.0m
400.0m
Volt [Lin]

200.0m
0
-200.0m
-400.0m
-600.0m
-800.0m
0 2.0m 4.0m 6.0m 8.0m 10.0m 12.0m 14.0m 16.0m 18.0m20.0m
Time [Lin]

Figure 87 Transient Responses of the Circuit and Pole/Zero Models

330 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Modeling Switched Capacitor Filters

-6.4792 lowpass9a
vp(out
-100.0
-200.0
Volt deg [Lin]

-300.0
-400.0
-500.0
-600.0
-700.0
-800.0
-6.4792 ninth.ac0
v(outp
-100.0
Volt deg [Lin]

-200.0
-300.0
-400.0
-500.0
-600.0
-700.0
-800.0
100.0 1.0k 10.0k 100.0k
Hertz [Log]

Figure 88 AC Analysis Responses of the Circuit and Pole/Zero Models

Modeling Switched Capacitor Filters

The following sections discuss these topics:
■
Switched Capacitor Network
■
Switched Capacitor Filter Example
■
Input File for Switched Capacitor Filter

Switched Capacitor Network

You can model a resistor as a capacitor and switch combination. The value of
the equivalent is proportional to the frequency of the switch, divided by
capacitance.
Construct a filter from MOSFETs and capacitors, where the filter characteristics
are a function of the switching frequency of the MOSFETs.

HSPICE® User Guide: Advanced Analog Simulation and Analysis 331

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Modeling Switched Capacitor Filters

To quickly determine the filter characteristics, use ideal switches (voltage-

controlled resistors), instead of MOSFETs. The resulting simulation speed-up
can be as great as 7 to 10 times faster than a circuit using MOSFETs.
To construct an RC network, the model uses a resistor and a capacitor, along
with a switched-capacitor equivalent network. The RCOUT node is the resistor/
capacitor output and VCROUT is the switched-capacitor output.
The GVCR1 and GVCR2 switches and the C3 capacitance model the resistor.
The following equation calculates the resistor value:
Tswitch
Res = --------------------
C3
where Tswitch is the period of the PHI1 and PHI2 pulses.

Figure 89 VCR1.SP Switched Capacitor RC Circuit

Switched Capacitor Filter Example

This example is a fifth-order elliptic, switched-capacitor filter. The passband is
0-1 kHz, with loss less than 0.05 dB. This results from the cascading models of
the switches. The resistance is1 ohm when the switch is closed and 100
Megohm when it is open. The E-element models op-amps as an ideal op-amp.

332 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Modeling Switched Capacitor Filters

This example provides the transient response of the filter for 1 kHz and 2 kHz
sinusoidal input signals.

Cd S6

Csh S5
Ce

-
In S1 Cs S4
+ Out
OP1
S2 S3

Figure 90 Linear Section

C2 S10

C4 S9

S1 C1 S4 Ca Cb
- S5 C3 S7
In -
+
OP1 + Out
S2 S3 OP2
S6 S8

C11

Figure 91 High_Q Biquad Section

HSPICE® User Guide: Advanced Analog Simulation and Analysis 333

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
Modeling Switched Capacitor Filters

C2 S11

S10
Ca C4 Cb
S1 S2 C1 S5
- S6 C3 S8 S12
In -
+
OP1 + Out
S3 S4 OP2
S7 S9

C11

Figure 92 Low_Q Biquad Section

Input File for Switched Capacitor Filter

This sample input file for a switched capacitor filter is located in the following
directory:
$installdir/demo/hspice/behave/swcap5.sp
This file also contains the following examples:
■
Sample and Hold
■
Linear Section
■
High_Q Biquad Section
■
Low_Q Biquad Section

334 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Chapter 14: Laplace Transform Modeling of Filters and Networks
Modeling Switched Capacitor Filters

1.50 00CAPE.TR0
V(OUT

3.0
Volt [Lin]

200.0H

-200.0H

-1.H
0 890.0H 1.280H 1.320.0H 2.580.0H 9.200H
Time [Lin]

Figure 93 Response to 1-kHz Sinusoidal Input

300.0 00CAPE.TR0
V(OUT

200.0
Volt [Lin]

100.0

-100.0

-200.0
0 890.0H 1.280H 1.320.0H 2.580.0H 9.200H
Time [Lin]

Figure 94 Response to 2-kHz Sinusoidal Input

HSPICE® User Guide: Advanced Analog Simulation and Analysis 335

N-2017.12
Chapter 14: Laplace Transform Modeling of Filters and Networks
References

References
[1] Williams, Arthur B., and Taylor, Fred J. Electronic Filter Design Handbook.
New York: McGraw-Hill, 1988, pp. 6-20 to 6-23.
[2] Nillson, James W. Electric Circuits, 4th Edition. Reading, Massachusetts:
Addison-Wesley, 1993.
[3] Edminister, Joseph A. Electric Circuits. New York: McGraw-Hill, 1965.
[4] Ghausi, Kelly, and M.S. “On the Effective Dominant Pole of the Distributed
RC Networks”, Jour. Franklin Inst., June 1965, pp. 417- 429.
[5] Elmore, W.C. and Sands, M. Electronics, National Nuclear Energy Series,
New York: McGraw-Hill, 1949.
[6] Pillage, L.T. and Rohrer, R.A. “Asymptotic Waveform Evaluation for Timing
Analysis”, IEEE Trans. CAD, Apr. 1990, pp. 352 - 366.
[7] Kuo, F. F. Network Analysis and Synthesis. John Wiley and Sons, 1966.
[8] Gregorian, Roubik & Temes, Gabor C. Analog MOS Integrated Circuits. J.
Wiley, 1986, page 354.

336 HSPICE® User Guide: Advanced Analog Simulation and Analysis

N-2017.12
 Index

A elements 264, 279

AC analysis flip-flop 270
circuit and pole/zero models 331 gates 267
algorithm LC oscillator 290
nonlinear perturbation 102 look-up tables 272
periodic AC 102 n-channel MOSFETs 269, 274
AM p-channel MOSFETs 269
behavioral 282 phase detector model 293
amplifier 18 phased locked loop 293
amplifier, IP3 22 ring oscillator 278
amplitude modulation (AM) 108 VCO model 288
AM-PM separation 108 VVCAP model 291
analog behavioral modeling 263
elements 279 broadband phase noise 102
modeling 266 broadband phase noise validation 102
analysis
circuit model 331
phase noise 97
C
pole/zero capacitance
model 331 voltage variable 291
time domain steady-state 51 capacitor
transfer function (.TF) 277 switched 331
circuits
asymptotic waveform evaluation 300, 323
AC analysis 331
autocorrelation function 225
model 329
autocorrelation function, jitter measurement 235 RC line 321
AWE See asymptotic waveform evaluation transient responses 330
clock source, random jitter 120
B CMOS GPS VCO 75
behavioral coefficients
741 op-amp 284 Laplace 313
amplitude modulator 282 transfer function 310
AND and NAND gates 267 Colpitts oscillator 69
BJTs 294 .command
CMOS inverter 276 .PRINT ENV 200
comparator 287 command
components 272 .PROBE ENV 200
controlled sources 266 commands
data sampler 283 PTDNOISE 251
differentiator 281 comparators 284, 287
D-Latch 268 comparing results 138
double-edge triggered flip-flop 270 complex poles and zeros 318

337
Index
D

component noise models 217 E

config file E Elements 267, 315–321
hspicerf 8
integrator application 281
configuration file 8 Laplace transform 301
example 11 NAND gate 267
configuration options SCALE parameter 325
flush_waveform 9 element
ground_floating_ node 9 analog behavioral 279
hier_delimiter 9 current controlled 267
html 9 voltage-controlled 267
integer_node 9
element noise models 217
max_waveform_size 9
Elmore delay 321
negative_td 9
.ENV statement 197
port_element_ voltage_ matchload 10
rcxt_divider 10 Envelope Analysis (ENV) 195
unit_atto 10 envelope simulation 195
v_supply 10 .ENVFFT 198
wildcard_left_range 10 .ENVFFT statement 198
wildcard_match_all 10 .ENVOSC 197
wildcard_match_one 10 .ENVOSC statement 197
wildcard_right_range 10 EPSMIN option 310
controlled sources 266, 267 example
critical frequency 302 configuration file 11
current
controlled
elements 267 F
Custom WaveView 7 F Elements 267
files
.hl# 189
D hspicerf 8
data .ls# 193
sampler, behavioral 283 .p2d# 193
DDL 266 .printhl# 189
delays .printls# 193
Elmore 321 .printss# 193
DELF element parameter 303 .ss# 193
demo files files, output 6
13 filters
demo files, RF 13 30 degree phase shift 307
demonstration files, RF 13 band reject 303
demonstration input files 13 low-pass 305
design transfer functions 319
stability 300 flags 8
time, reducing 300 flicker noise 214
differentiator, behavioral 281 flush_waveform configuration option 9
digital Fourier
files 267 integral 302
transfer function H(f) 302

338
 Index
G

transform 302 output syntax 188

FREQ .HBLSP 189
keyword 302 example 191
frequency input syntax 190
complex 301, 310 limitations 194
critical 302 output data files 189, 193
domain to time domain 299 output syntax 192
domain, transfer function 310 HBOSC analysis
maximum 303 Colpitts oscillator 69
resolution 303 VCO 75
response 301 HBXF
analysis 300 command 171
table 300 hier_delimiter configuration option 9
functions .hl# file 189
LAPLACE 300
hspicerf file 8
OPTIMIZE 265
hspicerf test 9
POLE 300
html configuration option 9

G
I
G Elements 267, 315–320
impulse response h(t) 302
AND gate 267
Laplace transform 301 input files demonstration 13
generating output 6 input files, demo examples 13
ground_floating_ node configuration option 9 integer_node configuration option 9
inverse Laplace transform 321
inverter lookup table 274
H
H Elements 267
Harmonic Balance (HB) 17
J
equations 30 jitter
errors 37 random, with clock source 120
output 34 jitter measurements, transient noise 235
warnings 37 jitter, random, clock source 120
.HB
for HBLIN 185 K
HB analysis
keywords
IP3 amplifier 22
FREQ 302
power amplifier 18
HB_GIBBS option 43
HBAC 133, 136 L
errors 144, 150, 249 Laplace
example 136 band-reject filter 303
output 141, 199 function 300, 302, 309, 313–317
output data files 143 low-pass filter 305
syntax 141 transfer function 300, 309
warnings 144, 150, 249 transform 299, 312
.HBLIN 182, 186 function call 301
limitations 184 inverse 321

339
Index
M

modeling 309 syntax 48

POLE (pole/zero) function 317 OPTIMIZE keyword 265
large-signal S parameter extraction 189 .OPTION HB_GIBBS 43
LC oscillator model 290 options
.ls# file 193 EPSMIN 310
oscillator
phase noise 97
M oscillator example 69
mA741 op-amp 284 oscillators
max_waveform_size configuration option 9 behavioral models 284
MAXF parameter 303 LC 290
.measure 200 VCO 288
.MEASURE ENV command 200 output
files 6
mixed mode simulation 266
mixed-signal simulation generating 6
See mixed mode restricting 12
models
behavioral 263 P
Monte Carlo, TRANNOISE method 225
.p2d# file 193
Pade approximation 327
N periodic AC algorithm 102
negative_td configuration option 9 periodic pime-dependent noise analysis 251
network, switched capacitor 331 phase
noise detector model 293
component models 217 locked loop, BJT model 294
element models 217 phase modulation (PM) 108
flicker 214 phase noise 97
.HBNOISE 155, 163 PHASENOISE algorithms 101
shot 216 PLL See phase locked loop
simulation 217, 218 PLL, jitter measurements 113
sources 210 POLE
thermal 212 function 300, 317–321
types 211 call 317
noise parameter extraction highpass filter 319
small-signal 189 low-pass filter 320
noise simulation 205 pole/zero
nonlinear elements 300 analysis 300
nonlinear perturbation algorithm 102 complex 318
Nyquist critical frequency 302 models 329, 330, 331
transfer function 317
transient modeling 300
O port_element_voltage_matchload configuration
op-amps option 10
behavioral models 284 power amplifier 18
mA741 model 284
power amplifier IP3 22
operators 312
power-dependent S parameter extraction 189
optimization
.PRINT ENV command 200
behavioral models 276, 278

340
 Index
R

.printhl# file 189 SN steady-state time domain analysis 51

.printls# file 193 SNAC
.printss# file 193 input syntax 144
.PROBE command 200 output data files 149
PTDNOISE SNFT 61
input syntax 253 SNNOISE 163
.MEASURE 256 input syntax 165
output file format 255 output data files 168
output syntax 254 SNOSC 90
overview 251 SNXF 175
syntax 253 source
PTDNOISE command 251 controlled 266, 267
.PZ statement 300 element types 267
spur (spurious) signals 100
.ss# file 193
R
statement 197, 198
random signals, characteristcs 206
.ENV 197
RC
statements
line modeling 321
.HBXF 171
rcxt_divider configuration option 10 .PTDNOISE 251
restricting output 12 steady state time domain analysis, Shooting
results 138 Newton 51
RF Stochastic Differential Equation, TRANNOISE
demo files 13 method 225
RFdemo files 13 strobejitter 256
switched capacitor 331, 332
S
S parameter extraction T
large-signal 189 .TF statement 277
power-dependent 181 thermal noise 212
small-signal 189
TIE 225
SCALE parameter 325
TIE, jitter mreasurement 235
SDE 225, 245 time
s-domain 309, 314 domain
separating AM-PM noise 108 to frequency domain 299
Shooting Newton maximum 303
driven phase frequency circuit example 52 resolution 302
overview 51 time domain steady state analysis 51
ring oscillator example 91 time interval error (TIE) 225
shot noise 216 TRANNOISE
simulation jitter measurements 225
reducing time 332 SDE method 245
skip_nrd_nrs configuration option transfer function
configuration options algebraic 265
skip_nrd_nrs 10 analysis 277
small-signal noise parameter extraction 189 coefficients 319
small-signal S parameter extraction 189 filters 319

341
Index
U

frequency domain 310 elements 267

general form 309, 317 oscillator 288
inverse Laplace transform 321 voltage source 287
reduced form 318 VVCAP model 291
voltage gain 313
transforms, Fourier 302
transient
W
modeling 300 warnings
MAXF ignored 303
transient modeling 299
waveform viewing 7
transient noise jitter measurements 235
WaveView 7
two-tone HB 136
wildcard uses 11
wildcard_left_range configuration option 10
U wildcard_match_all configuration option 10
unit_atto configuration option 10 wildcard_match_one configuration option 10
wildcard_right_range configuration option 10
V
v_supply configuration option 10 Y
VCO 75 Y parameter
VCO See voltage controlled oscillator line model 329
Verilog-A support 4 modeling 326
voltage
network 327
variable capacitance model 291
voltage-controlled

342