SlideShare a Scribd company logo
Martin Vlach
Chief Technologist AMS
July 2014
Mixed Signal Verification
Transistor to SoC
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Agenda
Mixed Signal Verification July 2014
 AMS Verification Landscape
 Verification vs. Design
 Issues in AMS Verification Modeling
 Summary
2
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
AMS VERIFICATION
LANDSCAPE
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
 It is the interactions that kill projects
— Simple: Connect “stuff wrong”
— Complex: Connect the “wrong stuff” (algorithms)
— Electrical: It’s all on the same silicon
 Project phases addressed by AMS Verification
— Design phase: RTL and schematics
— Power management: UPF for digital and analog
— Post-synthesis phase: Gates and schematics
— Post-layout: SDF and DSPF
 Human factors
— Understand each other: Do we “speak” the same tools
— Solve problems: Advanced debugging and diagnostics
— Integrated flows: Solving problems, not fighting tools
— Manage projects: Do we know the state of our project
Mixed Signal Verification Challenges
Mixed Signal Verification July 2014
4
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Four Kinds of AMS Verification
Mixed Signal Verification July 2014
5
 Functional Verification
— The task of verifying that digital logic and analog
input-output requirements are met
 Parametric Verification
— The task of verifying that numerical requirements are
met
 Implementation Verification
— The task of verifying that functional and parametric
requirements are met considering all the ways that
circuits can “go wrong” in an “analog way”
 Reliability Verification
— The task of verifying that requirements continue to
be met as prescribed by the reliability requirements.
Digital
Analog
Analog
Digital
Analog
Analog
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Shifting Landscape: Working Together
Mixed Signal Verification July 2014
6
 Simulation is the problem
 Simulators are the answer
 AMS Verification is the problem
 A common verification strategy
for Digital, Analog and Mixed
Blocks
 Predictable, repeatable, and
observable process
 Advanced debugging, diagnosis,
analysis, requirements tracing
Analog
Analog
Digital
Digital
GUI GUI
Analog
Analog
Digital
Digital
Common Environment
Debug, Diagnosis, Analysis, Req Tracing
Veloce
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
AMS Verification : Three Solutions
Mixed Signal Verification July 2014
7
 Event-Driven and Real Number Modeling
— Fastest, least accurate, big digital best choice.
— Questa, Questa ADMS
 Mixed-Signal (aka AMS) Modeling
— Fast, can be accurate, big analog best choice.
— Questa ADMS
 Digital/SPICE joined simulation
— Slowest, most accurate analog, easiest.
— Questa
— Eldo, Eldo Premier, ADiT
— Analog FastSPICE from BDA acquisition
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Putting AMS Together
Mixed Signal Verification July 2014
8
 Apart and together
— Verifying Digital and Analog when apart: building blocks
— Verifying Digital with Analog together: AMS IP
— Verifying Digital and Analog integration: SoC
— Adding power with UPF
 Testbenches
— SystemVerilog with UVM (-MS)
— SPICE or HDL-AMS
 Planning and management
— Expanding Verification Run Management to AMS
— Expanding UCDB to AMS
— Expanding application of ReqTracer to AMS
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
VERIFICATION
VS.
DESIGN
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
What Is Verification
Mixed Signal Verification July 2014
10
 Verification is the science (and art) of asking the question, “What
could possibly go wrong?” [Bryon Moyer]
 Verification in a colloquial sense
— Digital functional verification
— For analog circuits making sure, using SPICE simulation, that “all process
corners have been covered”
— For mixed signal circuits, verification seems to have the sense of
“running a time-domain simulation in a mixed-signal simulator”
 Verification in the formal sense used in Quality Management systems
and Functional Safety standards has these objectives:
— Evidence is provided that the implementation meets the requirements.
— Traceability is established between hardware requirements, the
implementation, and the verification procedures and results.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Verification vs. Design
Mixed Signal Verification July 2014
11
 In the formal sense of standards, design and verification
are separate and concurrent processes.
 Design engineers
— Are concerned with how to meet a specification (perhaps a
requirement, or a derived requirement).
— Think of how everything will “go right”.
— Do not question requirements and specifications, they implement
to them.
 Verification engineers
— Are concerned with whether a requirement has been met.
— Think of what can “go wrong”.
— Worry whether derived requirements correctly fulfill system
requirements.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Stakeholders - Technical
Mixed Signal Verification July 2014
12
 Digital verification engineer
— Understands verification
— Uses SV testbenches
— Relies on RTL delivered from design for digital blocks
— Uses very high level (TLM) models of analog blocks
– that reflect digital concerns
– that are derived from specifications in English
 Analog designer
— Understands analog design
— Uses schematics and SPICE
— Designs to specifications in English
— Does not trust behavioral models as they do not reflect complete
analog interactions observable at SPICE level
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Stakeholders - Management
Mixed Signal Verification July 2014
13
 Management needs
— Repeatable processes
— Clarity
— Predictability
— Consistency
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
CONCEPTS IN AMS
VERIFICATION MODELING
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Modeling Purposes
Mixed Signal Verification July 2014
15
 Implementation model
— RTL or gate level for digital.
— The schematic for analog.
 Physical model
— Refinement of the design on the way to fabrication. An implementation
model together with physical layout (timing, parasitics, etc.)
 Behavioral model
— Attempts to model behavior even in regions that will not be used.
— Often used as a reference model.
— Top-down design and verification.
 Verification model
— Models behavior only for specific purpose, and uses assertions to identify
out-of-spec use of the model.
— Bottom-up verification.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Model Levels
Mixed Signal Verification June 2014
16
 Level 0 - Empty: A model that literally has no information other than
the definition of the interface. Although counterintuitive, empty models can
serve an important role as placeholders during integration.
 Level 1 – Anchored load: The inputs may present a load to the
driving nodes, and the output is fixed at a typical value. The model can
include checking code, for example to make sure that correct levels are
established by biasing, and can already be useful in simulation to check that
inputs stay within design bounds.
 Level 2 – Feed-through: The signal levels at inputs are observed
and drive a related output value, although the full signal processing
functionality of the component is not represented. This may be used for
example to ensure that correct supplies are connected.
 Level 3 – Basic: At this modeling level, first order functionality of the
block is represented at one environment point, with input wiggles being
transmitted to related output wiggles.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Model Levels
Mixed Signal Verification June 2014
17
 Level 4 – Functional: The model is fully functional at one
environment point, and includes important second order effects.
Practically speaking, this is the last level of verification model that is
likely to be used in most projects.
 Level 5 – High Fidelity: Fully functional model at all
environment points. Only very sophisticated teams will find a need
for, and especially find beneficial return on investment, at this
modeling level. Maintaining such high fidelity models is very
expensive, and creating them in the first place and making sure that
the model actually reflects what has been built is technically very
difficult.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Signal Abstractions in AMS Verification
Mixed Signal Verification July 2014
 Abstractions for modeling a DUT real-life signal
— Logic
— Electrical
— Signal flow
— Real Number (RN) modeling
— Event Driven (ED) modeling
 Logic
— Well understood, digital flows
— Discrete time, event-driven simulation
 Electrical (special case of conservative)
— Well understood, analog flows
— Continuous time, DAEs, need analog kernel
— Coded in VHDL-AMS, Verilog-AMS, Verilog-A
 Signal flow
— Not used in verification, only at system/architecture exploration
— Continuous time, DAEs, need analog kernel
18
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Real Number Signal Abstraction
Mixed Signal Verification July 2014
 Real Number (RN) modeling
— Similar to signal flow in that information propagates from inputs
to outputs; no loading
— Discrete time, event-driven simulation
— Signal is represented by one real number
— Special case of Event Driven modeling
— Coupled with concepts of X (unknown) and Z (high impedance)
values
— Implementation:
– wreal
– Verilog-AMS 2.3 with Cadence extensions
– SV with wreal extensions
– VHDL real subtype, possibly resolved, user-defined
– System Verilog (SV) real variable (unresolved single driver)
– essentially identical to Verilog-AMS without the Cadence extensions
– SV User Defined Nettype (UDN) with a real type, possibly resolved
19
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Event Driven Signal Abstraction
Mixed Signal Verification July 2014
 Event Driven (ED) modeling
— ED modeling is a proper superset of RN modeling
— Signal is represented by multiple pieces of information, e.g.
– voltage/current/impedance
– frequency/phase/dutycycle
— Implementation:
– VHDL: composite type, usually record, possibly resolved
– SV: aggregate type, usually struct, possibly resolved, using the
User Defined Nettypes (UDNs) construct
20
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Connectivity
Mixed Signal Verification July 2014
 Mixed signal design
— Where digital and analog flows interact
— Not necessarily two simulation engines
 Mixed (digital) language (aka Mixed design in digital idiom)
— SV and VHDL, often for SV testbench and connectivity and VHDL Design
Under Test (DUT)
 Mixed abstraction
— Logic and electrical for mixed engine
— Logic and RN in digital engine
— Logic/electrical/RN/ED in the most general case
 Mixed kernel simulation
— Implies mixed abstraction: commonly electrical/logic or electrical/RN
 Connectivity: putting models together:
— Abstract connectivity (no representation known)
— Concrete connectivity (representation explicitly specified)
— Structure from schematic with netlisters
21
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Mixed Abstraction Issues
Mixed Signal Verification July 2014
 May arise in digital-only kernel simulations, always arise in
mixed-kernel
 Analog abstraction conversion
— Boundary element insertion is subject to discussion related to
accuracy and simulation speed tradeoff: where and how many?
 electrical - logic
 electrical - wreal
 wreal - logic
 electrical - wreal - logic
 Multiple types in general, including all abstractions and
languages
22
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Major AMS Flows
Mixed Signal Verification July 2014
23
 Testbenches with UVM for MS
— Stimulus: driving RN and ED signals from SV testbench
— Monitoring: reading ED and RN model values from SV testbench
 Low Power with UPF for MS
— Connecting the UPF power/ground with the analog world
– Explicit supplies, either in SPICE or as RN or ED
— Abstraction conversion (boundary element insertion) in the
presence of multiple power domain
– With level shifters, isolation cells, and retention cells inserted
– With direction signal connection between power domains
 Model Verification
— Not model equivalence checking as understood in digital
— Making sure the verification model corresponds to the transistor
implementation
— Use the same testbench to drive SPICE and RN/ED models
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Languages
Mixed Signal Verification July 2014
24
Language Strengths
Simulink (simulator) Architecture, signal flow
SystemVerilog(SV) Large complex SoCs, UVM testbench, DUTs, Emulation
Verilog Pre‐SV, basis for Verilog‐AMS
VHDL System through semiconductor, early HDL
SystemC System level
Verilog‐A Semiconductor, PDKs, behavioral analog (no digital)
Verilog‐AMS Digital‐analog behavioral modeling, semiconductor
VHDL‐AMS System through semiconductor, early HDL‐AMS
SystemC‐AMS System level
SV‐AMS Early stages of standard work
SPICE Transistor level
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Practical problems in AMS Verification
Mixed Signal Verification July 2014
25
 Have a transistor implementation of an analog IP block
 Have a model of that implementation (2k-3k digital pins)
 Convince me that the model(s) faithfully represent(s) the
circuit
— The full specifications of the circuit is modeled
— For the purpose to which the circuit will be put
– Do verification "in situ"
— That if I misuse the model (connect it incorrectly, connect it in
the wrong environment) that I will
– Find out that there is a bug
– Find out what the bug is
 Have an SoC testbech that uses the model
— Convince me that the testbench tests all the requirements
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Verification Modeling Best Practices
Mixed Signal Verification July 2014
26
 Executable specifications (reference and verification models) are used
in addition to English.
 Modeling technologies and styles are consistent with the verification
problem.
 AMS verification modeling of mixed signal and analog blocks includes
both the verification engineer and the analog designer.
 Continuous modeling effort is an integral part of the verification
process, not an afterthought or a one-time activity.
 Assertion Based Verification (ABV) is used in the MS models.
 Model verification is an integral part of the regression testing
environment.
 Formalized testbenches (SV?) are used in verifying the model against
the SPICE implementation.
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
SUMMARY
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential
Summary
Mixed Signal Verification July 2014
28
 Mixed signal verification is more than running a mixed-
signal simulations.
 Mixed signal verification methodology often requires
investment in mixed signal modeling.
— To catch bugs early
— To cope with specification changes
 Modeling of AMS blocks for SoC verification does not have
a one-size-fits-all answer.
— AMS modeling is difficult in the first place
— Knowing what not to model is harder
— Knowing whether everything that needs to be models in fact is
modeled is hardest - analog coverage
www.mentor.com
© 2014 Mentor Graphics Corp. Company Confidential

More Related Content

PDF
Presentation Verification & Validation
Elmar Selbach
 
DOC
Meganathan_Profile
Meganathan B
 
PPTX
Automotive Software Cost Estimation - The UCE Approach - Emmanuel Mary
Nesma
 
PPTX
Building products - A Nifty Approach
GuruprasadBhat21
 
DOCX
Chethan Updated Resume
Chethan H
 
PDF
EliteMindz: Who are we? Where do we serve ? What are our products & services?
Simerjeet Singh
 
PDF
Elite mindz introduction
Simerjeet Singh
 
DOCX
AlekyaNyalakonda_Resume
Alekya Nyalakonda
 
Presentation Verification & Validation
Elmar Selbach
 
Meganathan_Profile
Meganathan B
 
Automotive Software Cost Estimation - The UCE Approach - Emmanuel Mary
Nesma
 
Building products - A Nifty Approach
GuruprasadBhat21
 
Chethan Updated Resume
Chethan H
 
EliteMindz: Who are we? Where do we serve ? What are our products & services?
Simerjeet Singh
 
Elite mindz introduction
Simerjeet Singh
 
AlekyaNyalakonda_Resume
Alekya Nyalakonda
 

Similar to FAC14_Vlach.pdf (20)

PDF
CV-For-Thair 2
tjust80
 
DOCX
FazilShaikh Resume 13th january
fazilahmed sheikh
 
DOCX
FazilShaikh Resume 13th january
fazilahmed sheikh
 
PDF
An IoT-Based Smart Pothole Detection and Reporting System for Real-Time Road ...
katurechahak
 
PPTX
Neev Application Performance Management Services
Neev Technologies
 
DOC
Sainath_Resume
Sainath Tankasala
 
PDF
Surekha_haoop_exp
surekhakadi
 
PPT
4.4.2013 Software Quality - Regression Testing Automated and Manual - RFT/RQM
IBM Rational
 
PDF
Ensuring Design Quality in Mixed Signal IP
Arasan Chip Systems
 
PPTX
End to-End Monitoring for ITSM and DevOps
eG Innovations
 
PDF
A Practical Guide To Implementing Smart Process Applications
Sandy Kemsley
 
PDF
Bangalore based Test automation and Testing service Company - Oak Systems Pvt...
OAK Systems Pvt Ltd
 
PDF
Dallas Salesforce Women in Tech: Dreamforce Review '24.pdf
kellywood62
 
PDF
Deployment Automation for Hybrid Cloud and Multi-Platform Environments
IBM UrbanCode Products
 
PPTX
Model-Based Testing for Cypress
Curiosity Software Ireland
 
PPTX
Amtex Software Testing Presentation_v7
Manish Chugh
 
PDF
Resume
Shiv Dutta
 
KEY
Skillshare - From Noob to Tech CEO - nov 7th, 2011
Kareem Amin
 
PDF
Integrating AI in software quality in absence of a well-defined requirements
Nagarro
 
CV-For-Thair 2
tjust80
 
FazilShaikh Resume 13th january
fazilahmed sheikh
 
FazilShaikh Resume 13th january
fazilahmed sheikh
 
An IoT-Based Smart Pothole Detection and Reporting System for Real-Time Road ...
katurechahak
 
Neev Application Performance Management Services
Neev Technologies
 
Sainath_Resume
Sainath Tankasala
 
Surekha_haoop_exp
surekhakadi
 
4.4.2013 Software Quality - Regression Testing Automated and Manual - RFT/RQM
IBM Rational
 
Ensuring Design Quality in Mixed Signal IP
Arasan Chip Systems
 
End to-End Monitoring for ITSM and DevOps
eG Innovations
 
A Practical Guide To Implementing Smart Process Applications
Sandy Kemsley
 
Bangalore based Test automation and Testing service Company - Oak Systems Pvt...
OAK Systems Pvt Ltd
 
Dallas Salesforce Women in Tech: Dreamforce Review '24.pdf
kellywood62
 
Deployment Automation for Hybrid Cloud and Multi-Platform Environments
IBM UrbanCode Products
 
Model-Based Testing for Cypress
Curiosity Software Ireland
 
Amtex Software Testing Presentation_v7
Manish Chugh
 
Resume
Shiv Dutta
 
Skillshare - From Noob to Tech CEO - nov 7th, 2011
Kareem Amin
 
Integrating AI in software quality in absence of a well-defined requirements
Nagarro
 
Ad

More from SamHoney6 (14)

PDF
eetop.cn_UVM_REG_workshop.--------------pdf
SamHoney6
 
PDF
Cadence GenusTutorial------------ .pdf.pdf
SamHoney6
 
PDF
snug07_Verilog Gotchas for Verification.pdf
SamHoney6
 
PDF
04+ECETEMT092-+WDT+APB+UVM.pdf
SamHoney6
 
PDF
14-Bill-Tiffany-SigmaSense-VF2023.pdf
SamHoney6
 
PDF
RF-Transceiver.pdf
SamHoney6
 
PDF
inf5430_sv_randomization.pdf
SamHoney6
 
PDF
vip-shielding.pdf
SamHoney6
 
PDF
UVM_TB_20220621_slides-1.pdf
SamHoney6
 
PDF
Sva.pdf
SamHoney6
 
PDF
Web Template Mechanisms in SOC Verification - DVCon.pdf
SamHoney6
 
PPT
cupdf.com_chapter-11-system-level-verification-issues-the-importance-of-verif...
SamHoney6
 
PDF
UVM-based RISC-V processor Verification Paltform ---.pdf
SamHoney6
 
PDF
SVA Advanced Topics- SVAUnit and Assertions for Introduction to SystemVerilog...
SamHoney6
 
eetop.cn_UVM_REG_workshop.--------------pdf
SamHoney6
 
Cadence GenusTutorial------------ .pdf.pdf
SamHoney6
 
snug07_Verilog Gotchas for Verification.pdf
SamHoney6
 
04+ECETEMT092-+WDT+APB+UVM.pdf
SamHoney6
 
14-Bill-Tiffany-SigmaSense-VF2023.pdf
SamHoney6
 
RF-Transceiver.pdf
SamHoney6
 
inf5430_sv_randomization.pdf
SamHoney6
 
vip-shielding.pdf
SamHoney6
 
UVM_TB_20220621_slides-1.pdf
SamHoney6
 
Sva.pdf
SamHoney6
 
Web Template Mechanisms in SOC Verification - DVCon.pdf
SamHoney6
 
cupdf.com_chapter-11-system-level-verification-issues-the-importance-of-verif...
SamHoney6
 
UVM-based RISC-V processor Verification Paltform ---.pdf
SamHoney6
 
SVA Advanced Topics- SVAUnit and Assertions for Introduction to SystemVerilog...
SamHoney6
 
Ad

Recently uploaded (20)

PPTX
Design & Thinking for Engineering graduates
NEELAMRAWAT48
 
PPTX
Engagement for marriage life ethics b.pptx
SyedBabar19
 
PPTX
Modern_Dhaka_Apartment_Interior_Design.pptx
hasansarkeraidt
 
PPTX
designing in footwear- exploring the art and science behind shoe design
madhuvidya7
 
PDF
Top 10 UI/UX Design Agencies in Dubai Shaping Digital Experiences
Tenet UI UX
 
DOCX
Personalized Jewellery Guide: Engraved Rings, Initial Necklaces & Birthstones...
Dishis jewels
 
PPTX
Economics Reforms of Alauddin Khilji.pptx
Kritisharma797381
 
PPTX
The Satavahanas.pptx....,......,........
Kritisharma797381
 
PPTX
UCSP-Quarter 1-Week 6-Powerpoint Presentation
EmyMaquiling1
 
PDF
SS27 Environments & Design Trend Book Peclers Paris
Peclers Paris
 
PPTX
United Nation - CoUnited Nation - CoUnited Nation - Copy (2).pptx
mangalindanjerremyjh
 
PPTX
Mini-Project description of design of expert project
peter349484
 
PDF
neurons-1hhdbbbsjndjnnbbdjndnnndj30501.pdf
Rahul954607
 
PPTX
Blended Wing Body y jet engines Aircrafts.pptx
anshul9051
 
PDF
Home_Decor_Presentation and idiea with decor
sp1357556
 
PDF
Garage_Aluminium_Doors_PresenGarage Aluminium Doorstation.pdf
Royal Matrixs
 
PDF
Fashion project1 kebaya reimagined slideshow
reysultane
 
PPTX
Web Design: Enhancing User Experience & Brand Value
ashokmakwana0303
 
PPTX
Riverfront Development_nashikcity_landscape
aditikoshley2
 
PDF
Unlimited G+12 Dubai DM exam questions for contractor
saniyashaik2089
 
Design & Thinking for Engineering graduates
NEELAMRAWAT48
 
Engagement for marriage life ethics b.pptx
SyedBabar19
 
Modern_Dhaka_Apartment_Interior_Design.pptx
hasansarkeraidt
 
designing in footwear- exploring the art and science behind shoe design
madhuvidya7
 
Top 10 UI/UX Design Agencies in Dubai Shaping Digital Experiences
Tenet UI UX
 
Personalized Jewellery Guide: Engraved Rings, Initial Necklaces & Birthstones...
Dishis jewels
 
Economics Reforms of Alauddin Khilji.pptx
Kritisharma797381
 
The Satavahanas.pptx....,......,........
Kritisharma797381
 
UCSP-Quarter 1-Week 6-Powerpoint Presentation
EmyMaquiling1
 
SS27 Environments & Design Trend Book Peclers Paris
Peclers Paris
 
United Nation - CoUnited Nation - CoUnited Nation - Copy (2).pptx
mangalindanjerremyjh
 
Mini-Project description of design of expert project
peter349484
 
neurons-1hhdbbbsjndjnnbbdjndnnndj30501.pdf
Rahul954607
 
Blended Wing Body y jet engines Aircrafts.pptx
anshul9051
 
Home_Decor_Presentation and idiea with decor
sp1357556
 
Garage_Aluminium_Doors_PresenGarage Aluminium Doorstation.pdf
Royal Matrixs
 
Fashion project1 kebaya reimagined slideshow
reysultane
 
Web Design: Enhancing User Experience & Brand Value
ashokmakwana0303
 
Riverfront Development_nashikcity_landscape
aditikoshley2
 
Unlimited G+12 Dubai DM exam questions for contractor
saniyashaik2089
 

FAC14_Vlach.pdf

  • 1. Martin Vlach Chief Technologist AMS July 2014 Mixed Signal Verification Transistor to SoC
  • 2. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Agenda Mixed Signal Verification July 2014  AMS Verification Landscape  Verification vs. Design  Issues in AMS Verification Modeling  Summary 2
  • 3. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential AMS VERIFICATION LANDSCAPE
  • 4. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential  It is the interactions that kill projects — Simple: Connect “stuff wrong” — Complex: Connect the “wrong stuff” (algorithms) — Electrical: It’s all on the same silicon  Project phases addressed by AMS Verification — Design phase: RTL and schematics — Power management: UPF for digital and analog — Post-synthesis phase: Gates and schematics — Post-layout: SDF and DSPF  Human factors — Understand each other: Do we “speak” the same tools — Solve problems: Advanced debugging and diagnostics — Integrated flows: Solving problems, not fighting tools — Manage projects: Do we know the state of our project Mixed Signal Verification Challenges Mixed Signal Verification July 2014 4
  • 5. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Four Kinds of AMS Verification Mixed Signal Verification July 2014 5  Functional Verification — The task of verifying that digital logic and analog input-output requirements are met  Parametric Verification — The task of verifying that numerical requirements are met  Implementation Verification — The task of verifying that functional and parametric requirements are met considering all the ways that circuits can “go wrong” in an “analog way”  Reliability Verification — The task of verifying that requirements continue to be met as prescribed by the reliability requirements. Digital Analog Analog Digital Analog Analog
  • 6. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Shifting Landscape: Working Together Mixed Signal Verification July 2014 6  Simulation is the problem  Simulators are the answer  AMS Verification is the problem  A common verification strategy for Digital, Analog and Mixed Blocks  Predictable, repeatable, and observable process  Advanced debugging, diagnosis, analysis, requirements tracing Analog Analog Digital Digital GUI GUI Analog Analog Digital Digital Common Environment Debug, Diagnosis, Analysis, Req Tracing Veloce
  • 7. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential AMS Verification : Three Solutions Mixed Signal Verification July 2014 7  Event-Driven and Real Number Modeling — Fastest, least accurate, big digital best choice. — Questa, Questa ADMS  Mixed-Signal (aka AMS) Modeling — Fast, can be accurate, big analog best choice. — Questa ADMS  Digital/SPICE joined simulation — Slowest, most accurate analog, easiest. — Questa — Eldo, Eldo Premier, ADiT — Analog FastSPICE from BDA acquisition
  • 8. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Putting AMS Together Mixed Signal Verification July 2014 8  Apart and together — Verifying Digital and Analog when apart: building blocks — Verifying Digital with Analog together: AMS IP — Verifying Digital and Analog integration: SoC — Adding power with UPF  Testbenches — SystemVerilog with UVM (-MS) — SPICE or HDL-AMS  Planning and management — Expanding Verification Run Management to AMS — Expanding UCDB to AMS — Expanding application of ReqTracer to AMS
  • 9. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential VERIFICATION VS. DESIGN
  • 10. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential What Is Verification Mixed Signal Verification July 2014 10  Verification is the science (and art) of asking the question, “What could possibly go wrong?” [Bryon Moyer]  Verification in a colloquial sense — Digital functional verification — For analog circuits making sure, using SPICE simulation, that “all process corners have been covered” — For mixed signal circuits, verification seems to have the sense of “running a time-domain simulation in a mixed-signal simulator”  Verification in the formal sense used in Quality Management systems and Functional Safety standards has these objectives: — Evidence is provided that the implementation meets the requirements. — Traceability is established between hardware requirements, the implementation, and the verification procedures and results.
  • 11. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Verification vs. Design Mixed Signal Verification July 2014 11  In the formal sense of standards, design and verification are separate and concurrent processes.  Design engineers — Are concerned with how to meet a specification (perhaps a requirement, or a derived requirement). — Think of how everything will “go right”. — Do not question requirements and specifications, they implement to them.  Verification engineers — Are concerned with whether a requirement has been met. — Think of what can “go wrong”. — Worry whether derived requirements correctly fulfill system requirements.
  • 12. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Stakeholders - Technical Mixed Signal Verification July 2014 12  Digital verification engineer — Understands verification — Uses SV testbenches — Relies on RTL delivered from design for digital blocks — Uses very high level (TLM) models of analog blocks – that reflect digital concerns – that are derived from specifications in English  Analog designer — Understands analog design — Uses schematics and SPICE — Designs to specifications in English — Does not trust behavioral models as they do not reflect complete analog interactions observable at SPICE level
  • 13. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Stakeholders - Management Mixed Signal Verification July 2014 13  Management needs — Repeatable processes — Clarity — Predictability — Consistency
  • 14. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential CONCEPTS IN AMS VERIFICATION MODELING
  • 15. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Modeling Purposes Mixed Signal Verification July 2014 15  Implementation model — RTL or gate level for digital. — The schematic for analog.  Physical model — Refinement of the design on the way to fabrication. An implementation model together with physical layout (timing, parasitics, etc.)  Behavioral model — Attempts to model behavior even in regions that will not be used. — Often used as a reference model. — Top-down design and verification.  Verification model — Models behavior only for specific purpose, and uses assertions to identify out-of-spec use of the model. — Bottom-up verification.
  • 16. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Model Levels Mixed Signal Verification June 2014 16  Level 0 - Empty: A model that literally has no information other than the definition of the interface. Although counterintuitive, empty models can serve an important role as placeholders during integration.  Level 1 – Anchored load: The inputs may present a load to the driving nodes, and the output is fixed at a typical value. The model can include checking code, for example to make sure that correct levels are established by biasing, and can already be useful in simulation to check that inputs stay within design bounds.  Level 2 – Feed-through: The signal levels at inputs are observed and drive a related output value, although the full signal processing functionality of the component is not represented. This may be used for example to ensure that correct supplies are connected.  Level 3 – Basic: At this modeling level, first order functionality of the block is represented at one environment point, with input wiggles being transmitted to related output wiggles.
  • 17. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Model Levels Mixed Signal Verification June 2014 17  Level 4 – Functional: The model is fully functional at one environment point, and includes important second order effects. Practically speaking, this is the last level of verification model that is likely to be used in most projects.  Level 5 – High Fidelity: Fully functional model at all environment points. Only very sophisticated teams will find a need for, and especially find beneficial return on investment, at this modeling level. Maintaining such high fidelity models is very expensive, and creating them in the first place and making sure that the model actually reflects what has been built is technically very difficult.
  • 18. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Signal Abstractions in AMS Verification Mixed Signal Verification July 2014  Abstractions for modeling a DUT real-life signal — Logic — Electrical — Signal flow — Real Number (RN) modeling — Event Driven (ED) modeling  Logic — Well understood, digital flows — Discrete time, event-driven simulation  Electrical (special case of conservative) — Well understood, analog flows — Continuous time, DAEs, need analog kernel — Coded in VHDL-AMS, Verilog-AMS, Verilog-A  Signal flow — Not used in verification, only at system/architecture exploration — Continuous time, DAEs, need analog kernel 18
  • 19. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Real Number Signal Abstraction Mixed Signal Verification July 2014  Real Number (RN) modeling — Similar to signal flow in that information propagates from inputs to outputs; no loading — Discrete time, event-driven simulation — Signal is represented by one real number — Special case of Event Driven modeling — Coupled with concepts of X (unknown) and Z (high impedance) values — Implementation: – wreal – Verilog-AMS 2.3 with Cadence extensions – SV with wreal extensions – VHDL real subtype, possibly resolved, user-defined – System Verilog (SV) real variable (unresolved single driver) – essentially identical to Verilog-AMS without the Cadence extensions – SV User Defined Nettype (UDN) with a real type, possibly resolved 19
  • 20. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Event Driven Signal Abstraction Mixed Signal Verification July 2014  Event Driven (ED) modeling — ED modeling is a proper superset of RN modeling — Signal is represented by multiple pieces of information, e.g. – voltage/current/impedance – frequency/phase/dutycycle — Implementation: – VHDL: composite type, usually record, possibly resolved – SV: aggregate type, usually struct, possibly resolved, using the User Defined Nettypes (UDNs) construct 20
  • 21. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Connectivity Mixed Signal Verification July 2014  Mixed signal design — Where digital and analog flows interact — Not necessarily two simulation engines  Mixed (digital) language (aka Mixed design in digital idiom) — SV and VHDL, often for SV testbench and connectivity and VHDL Design Under Test (DUT)  Mixed abstraction — Logic and electrical for mixed engine — Logic and RN in digital engine — Logic/electrical/RN/ED in the most general case  Mixed kernel simulation — Implies mixed abstraction: commonly electrical/logic or electrical/RN  Connectivity: putting models together: — Abstract connectivity (no representation known) — Concrete connectivity (representation explicitly specified) — Structure from schematic with netlisters 21
  • 22. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Mixed Abstraction Issues Mixed Signal Verification July 2014  May arise in digital-only kernel simulations, always arise in mixed-kernel  Analog abstraction conversion — Boundary element insertion is subject to discussion related to accuracy and simulation speed tradeoff: where and how many?  electrical - logic  electrical - wreal  wreal - logic  electrical - wreal - logic  Multiple types in general, including all abstractions and languages 22
  • 23. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Major AMS Flows Mixed Signal Verification July 2014 23  Testbenches with UVM for MS — Stimulus: driving RN and ED signals from SV testbench — Monitoring: reading ED and RN model values from SV testbench  Low Power with UPF for MS — Connecting the UPF power/ground with the analog world – Explicit supplies, either in SPICE or as RN or ED — Abstraction conversion (boundary element insertion) in the presence of multiple power domain – With level shifters, isolation cells, and retention cells inserted – With direction signal connection between power domains  Model Verification — Not model equivalence checking as understood in digital — Making sure the verification model corresponds to the transistor implementation — Use the same testbench to drive SPICE and RN/ED models
  • 24. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Languages Mixed Signal Verification July 2014 24 Language Strengths Simulink (simulator) Architecture, signal flow SystemVerilog(SV) Large complex SoCs, UVM testbench, DUTs, Emulation Verilog Pre‐SV, basis for Verilog‐AMS VHDL System through semiconductor, early HDL SystemC System level Verilog‐A Semiconductor, PDKs, behavioral analog (no digital) Verilog‐AMS Digital‐analog behavioral modeling, semiconductor VHDL‐AMS System through semiconductor, early HDL‐AMS SystemC‐AMS System level SV‐AMS Early stages of standard work SPICE Transistor level
  • 25. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Practical problems in AMS Verification Mixed Signal Verification July 2014 25  Have a transistor implementation of an analog IP block  Have a model of that implementation (2k-3k digital pins)  Convince me that the model(s) faithfully represent(s) the circuit — The full specifications of the circuit is modeled — For the purpose to which the circuit will be put – Do verification "in situ" — That if I misuse the model (connect it incorrectly, connect it in the wrong environment) that I will – Find out that there is a bug – Find out what the bug is  Have an SoC testbech that uses the model — Convince me that the testbench tests all the requirements
  • 26. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Verification Modeling Best Practices Mixed Signal Verification July 2014 26  Executable specifications (reference and verification models) are used in addition to English.  Modeling technologies and styles are consistent with the verification problem.  AMS verification modeling of mixed signal and analog blocks includes both the verification engineer and the analog designer.  Continuous modeling effort is an integral part of the verification process, not an afterthought or a one-time activity.  Assertion Based Verification (ABV) is used in the MS models.  Model verification is an integral part of the regression testing environment.  Formalized testbenches (SV?) are used in verifying the model against the SPICE implementation.
  • 27. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential SUMMARY
  • 28. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential Summary Mixed Signal Verification July 2014 28  Mixed signal verification is more than running a mixed- signal simulations.  Mixed signal verification methodology often requires investment in mixed signal modeling. — To catch bugs early — To cope with specification changes  Modeling of AMS blocks for SoC verification does not have a one-size-fits-all answer. — AMS modeling is difficult in the first place — Knowing what not to model is harder — Knowing whether everything that needs to be models in fact is modeled is hardest - analog coverage
  • 29. www.mentor.com © 2014 Mentor Graphics Corp. Company Confidential