Introduction to ISDE

Introduction to ISDE

Lloyd MassengillInstitute for Space and Defense Electronics

Vanderbilt UniversityNashville, Tennessee, USA, 37235

Vanderbilt Engineering

Massengill – ISDE Introduction 2DTRA 6.1 Kickoff – 5/08

Vanderbilt University Home of the Commodores (and the Radiation Effects Research Group and ISDE)

Located in Nashville, TN Private Institution ~11,000 students

Undergraduate: 6,532

Graduate/professional: 5,315

School of Engineering: 1,305 Engineering, Arts &

Sciences, Medicine, Nursing, Law, Business, Education, Music, Divinity

Degrees in 2007Baccalaureate: 1,468

MS: 1,062

PhD: 498



Vanderbilt Radiation Effects Program

30 graduate students A few undergraduate students Open access Basic research and support of ISDE

engineering tasks Training ground for rad-effects engineers

14 full time engineers 2 support staff ITAR compliant Support specific radiation effects engineering

needs in government and industry

Radiation Effects Research (RER) Group

Institute for Space and Defense Electronics (ISDE)

World’s largest university-based radiation effects program

10 faculty with extensive expertise in radiation-effects Beowulf supercomputing cluster Custom software codes EDA tools from multiple commercial vendors Multi-million $ aggregate annual funding Test and characterization capabilities and partnerships



DTRA-supported Grad Student “Product” Examples

> 25 peer-reviewed publications in 2007 under DTRA/RHM support > 35 presentations in 2007 under DTRA/RHM support 13 presentations accepted for IEEE NSREC 2008 with DTRA/RHM

credit line >8 DTRA-supported graduate student degrees awarded last two years



What is ISDE?

ISDE is a contract engineering unit of Vanderbilt University created to bring world-class support of space and DoD mission needs through radiation effects analysis and rad-hard design

ISDE brings several decades of “academic” resources/expertise and “real-world” engineering to bear on system-driven needs

ISDE provides: Government and industry radiation-effects resource

Modeling and simulation: RHTCAD, RHEDADesign support: radiation models, RHBDTechnology support: assessment, characterizationSystem support: systems engineering

Flexible staffing driven by project needsFacultyGraduate studentsProfessional engineering staff

ISDE Particulars: Established as a unit of Vanderbilt University: 1 Jan 2003 Academic staff: 10 faculty / ~30 graduate students Full-time engineering staff: 14 Support staff: 2



ISDE Capabilities

Support the design and analysis of radiation-hardened electronics Supply radiation effects models, design tools, and simulation services Provide engineering services for technology insertion and transfer Develop radiation hardness assurance test methods Address system-specific problems related to radiation effects Provide training to the community Retain a radiation effect “SWAT” team Reality training for future radiation effects “experts” (aka grad students)



Sampling of Current Projects

• U.S. Navy Trident II Life Extension (Draper prime)• Honeywell SOI-IV, TI BiCom 1.5, and Intersil EBHF technologies

• DTRA Radiation Hardened Microelectronics• IBM 9SF 90nm, TI 65 nm

• DARPA/DTRA Radiation Hardened by Design (Boeing prime)• IBM 8SF 130nm and 9SF 90 nm CMOS – Trusted Foundry

• NASA Electronic Parts & Packaging Program (NASA/GSFC)• IBM: 5HP, 8HP, 9SF 90nm, TI: 65 nm, 45 nm

• NASA Extreme Environment Electronics (Ga Tech prime)• IBM 5AM SiGe and BAE 150 nm CMOS

• CREME Monte Carlo (NASA MSFC/RHESE)• Aging of Electronics (U.S. Navy DTO/Lockheed-Martin)• U.S. Air Force Minuteman Technology Readiness• BAE SEU-Hardened SRAMs (BAE prime)• SEE Charge Collection Signatures at 90nm (and below) (ANT/IBM prime)• Virtual Irradiation Simulator Development (Air Force/AEDC/PKP)• Integrated Multi-scale Modeling of Molecular Computing Devices (DOE)• Substrate Charge Collection Studies (MEMC)• CFDRC TCAD Tool Development (DTRA SBIR and NASA STTR)• Lynguent Compact Model Development (DTRA SBIR)• SEU Analysis (Medtronic)• GaN HEMT/amplifier simulation (Lockheed Martin)• Radiation Effects on Emerging Electronic Materials and Devices (AFOSR/MURI)• Design for Reliability Initiative for Future Technologies (AFOSR/MURI through

UCSB)• DTRA Basis Research Efforts (three 6-1 grants)



USN D5LE Modeling Activities

AMS- Custom DevelopmentPDK DevelopmentEBHF – 5 Design-fab-eval cycles supportedSOI-IV – 5 Design-fab-eval cycles supportedBicom 1.5 – 2 Design-fab-eval cycles supported

DigitalIBISStandard Cell library validationSSI –SOI-IV & SOI-V

DiscreteActivesPassives

New Electrical Model CreationMagampPower MOSFET

Design Community Support (remote & local)Bugzilla – over 90 bugs reported, analyzed, & closedApp-notesModel inventoryTutorialsDesigner Interface meetings



USN D5LE Model Completion Summary

937 model files tested/calibrated/delivered to NEPL database 757 of these are ISDE custom developed and calibrated Over 100-million calibration simulations performed Significant support, training, design, simulation activities



USN D5LE Model Completion Summary

A few milestones: 45 major model releases/updates since Jan 2006 Complete PDK radiation models for EBHF, SOI-IV, BiCom Complete electrical, dose-rate, and degraded / corner models for all accepted program

parts Degraded parameter guide and corner models released PCIC macro, micro, design, simulation support – identified a feedback path design

enhancement to correct out-of-spec recovery time Enhanced macro models to include high-fidelity transient response (based on user

request) New MOSFET electrical models developed to the fill vendor gaps Developed and designed 8 test chips for program model calibration and verification Implemented an online community model support and feedback process Model training and designer interface meetings General ELDO training and aid



The Vandy to ISDE Connection

Vanderbilt has a comprehensive radiation effects analysis program to support DOD and commercial needs

Physics investigations – NASA/GSFC, NASA/MSFC, AFOSR MURI, DTRA 6.1 support – Vandy academic

Response mechanisms investigations – DTRA RHM, NASA, Navy support – Vandy academic / ISDE

RHBD development – DARPA RHBD and DTRA RHM support – ISDE



“Applied” Side of the Single Event Program

Through DTRA, DARPA, and NASA support, Vanderbilt has been investigating single-event mechanisms, circuit responses, hardening techniques, and rad-hard design from submicron to sub-100nm IC technology nodes

General Observations: Moore’s law complicates the testing, simulation, and analysis of all radiation

effects, especially single-events and soft error-rates The 250nm technology node was a watershed for the microelectronics

reliability community (especially those ‘radiation-concerned’). At 100-nm scale:Circuits that “should” be SEE hard are proving not to beCommercial ICs are showing alarming vulnerabilities to ground-based SEE

environmentsUnexpected SEE vulnerabilities (e.g. protons) have appeared

Why? Single events can no longer be considered localized, time-isolated, average

energy phenomena The ‘region of influence’ of an ion strike extends far beyond a single circuit ‘bit’

- spatially, logically, and temporally



“Applied” Side of the Single Event Program

Through DTRA, DARPA, and NASA support, Vanderbilt has been investigating single-event mechanisms, circuit responses, hardening techniques, and rad-hard design from submicron to sub-100nm IC technology nodes

General Observations: Moore’s law complicates the testing, simulation, and analysis of all radiation

effects, especially single-events and soft error-rates The 250nm technology node was a watershed for the microelectronics

reliability community (especially those ‘radiation-concerned’). At 100-nm scale:Circuits that “should” be SEE hard are proving not to beCommercial ICs are showing alarming vulnerabilities to ground-based SEE

environmentsUnexpected SEE vulnerabilities (e.g. protons) have appeared

Why? Single events can no longer be considered localized, time-isolated, average

energy phenomena The ‘region of influence’ of an ion strike extends far beyond a single circuit ‘bit’

- spatially, logically, and temporally

Heuristic approaches to IC hardening are failing

Failure (upset rate) predictions are failing

Comprehensive radiation effects modeling, incorporating a priori physics, is an essential part of mission-critical hardness assurance

Heuristic approaches to IC hardening are failing

Failure (upset rate) predictions are failing

Comprehensive radiation effects modeling, incorporating a priori physics, is an essential part of mission-critical hardness assurance



Example of VU Basic Research to ISDE Application to Community Tech Transfer:A “Real World” Problem

Baze broadbeam testing (Feb 07) revealed: 90nm RHBD DICE latches are hyper-sensitive to longitudinal-axis angular SE strikes

Upset saturated cross-sections approach unhardened designs Results do not follow conventional cos() charge collection rules

Baze broadbeam testing (Feb 07) revealed: 90nm RHBD DICE latches are hyper-sensitive to longitudinal-axis angular SE strikes

Upset saturated cross-sections approach unhardened designs Results do not follow conventional cos() charge collection rules

DICE 9SF shift register SEU test data

1E-11

1E-10

1E-9

1E-8

1E-7

0 20 40 60 80 100 120 140

LET (MeV/mg/cm2)

Cross Section (cm

-2)

SF-1-0deg-1111SF-1-0deg-0000SF-1-0deg-1010SF-1-0deg-1100

SF-3-0deg-1010SF-3-60degA-1010SF-3-60degB-1010WeibullSF-2-0deg-1010

60deg, orthogonal to rails

60deg, longitudinal to rails

Cro

ss S

ecti

on

(cm

2)

LET (MeV/mg/cm2)



“Real World” Issue

Issue: Boeing RHBD Phase 1.5 90nm DICE V1 latch did not meet SEE on-orbit error-

rate goals (< 1E-10 E/BD) based on broadbeam error data and CREME96 rate calculations

Cause: Phase 1.5 TCAD research work identified charge sharing as error mechanism

Complication: CREME96 (and other space error-rate codes)

do not properly handle layout-dependent effects (e.g. charge sharing) and

can significantly mis-predict error rates (by orders of magnitude) Therefore: unclear if DICE V1 or V2 on-orbit error rates, calculated for RHBD,

are accurate or dubious predictions



Resolution Strategy

VU “basic research” tools: Vanderbilt-ISDE has performed comprehensive TCAD analysis of SEE

mechanisms in sub-100nm technologies:

uncovered the importance of charge sharing

identified critical circuit node pairs (supported in part by DTRA/RHM, DARPA RHBD, NRL Albany Nanotech)

Vanderbilt-ISDE has developed a Monte-Carlo-based error-rate modeling technique that

operates from first principles physics for ion energy deposition – “virtual irradiation”

does not apply conventional error-rate assumptions

(supported in part by NASA/GSFC and DTRA)

Task Plan: Vanderbilt-ISDE was asked by the RHBD program to apply this technique to the

Phase 1.5 90nm DICE V2 latch in order to calculate a more accurate on-orbit error-rate expectation



Mixed-Mode TCAD DICE Setup

Calibrated 620/80 PMOS devices constructed in TCAD using ISDE physical description of the IBM 9SF FEOL technology

Calibrated 280/80 NMOS BSIM3 devices constructed in DESSIS-SPICE for pull-down loading



MRED Solid Modeling Component Setup

The solid model serves as the foundation for the radiation transport and calorimetry component of the analysis

Use GDSII layout information to generate an extruded model of the 9SF Latch Each layer is assigned an accurate compositional description – chemical

stoichiometry and density

Substrate, Active, and PolySubstrate, Active, and PolyOnlyOnly Substrate, Metallization,Substrate, Metallization,

and Passivation Shownand Passivation Shown



MRED/SPICE Interface

This project required the first application of the MRED-Spice coupling concept.

For each particle that strikes a sensitive volume, a Spice simulation is launched.

Each transistor’s collected charge is converted to a current pulse and directed to the appropriate node during run-time.

Q(TXQ(TXijij))

MRED MRED EventEventjj

TX1TX1

TX2TX2....TXnTXn

SPICESPICE(Circuit Template)(Circuit Template)

%I%I11

%I%I22

%I..%I..%I%Inn

Q DD

CLK CLK

D

CLK

QQPRECLR

FF1 FF2Irradiate FF1 at a random time and Irradiate FF1 at a random time and watch for an upset clocked out of FF2.watch for an upset clocked out of FF2.This process was repeated over This process was repeated over 100,000 times for the final simulation 100,000 times for the final simulation set.set.



Calibration to Broadbeam Data

Tilt

Roll

Tilt

Roll

0 10 20 30 40 50 60 7010-12

10-11

10-10

10-9

10-8

10-7

Simulation (50 MHz)Data (25<f<50 MHz)

Simulation 0o Tilt, 0o Roll Experiment 0o Tilt, 0o Roll

LET (MeVcm2/mg)

0 10 20 30 40 50 60 7010-12

10-11

10-10

10-9

10-8

10-7

Simulation 60o Tilt, 0o Roll Experiment 60o Tilt, 0o Roll Simulation 60o Tilt, 90o Roll Experiment 60o Tilt, 90o Roll

LET (MeVcm2/mg)

Simulation (50 MHz)Data (25<f<50 MHz)

Best agreement between model and experiment is with the highest cross sections and lowest LET – rate dominating



SEU Rate Prediction

To perform the rate prediction, the beam-calibrated model is modified to:

Mimic the isotropic environment and sample appropriately from each spectrum (z=1,z=2,z=3,etc.)

Events are weighted to the relative abundance in the overall spectrum. This methodology has been tested extensively and proven valid.

The calculated rate is 1.7 +/- 0.2 x 10-8 error/bit-day (the error bar is due to counting uncertainty only)

Most errors occurred at grazing incidence ( >60 degrees ) Began observing errors regularly around Z = 12 (Mg, max LET 10 MeV-

cm2/mg)

Tech Transfer: Based on Vandy analyses, improved V3 DICE latches have been designed

and fabbed by Boeing as part of the RHBD Phase 2.0 program Results on charge sharing, angular effects, well collapse, and MRED upset

rate modeling have been briefed to the community at NSREC, IRPS, GOMAC…



The “Big Picture”

Technology

Requirements

FunctionalDesign

ArchitectureDesign

Library ModuleDesign

Device Design / Layout

Qualification

FunctionalVerification

ArchitectureVerification

Library Validation

Component Response

Design Flow

Qualification Flow

M&S Enabled ASIC D,T,&Q

Rad-Aware EDA Virtual Irradiation

Targeted Radiation Testingfor M&S Support

3DMixed -Mode

TCAD

First PrincipleRadiation Physics

(MRED)

Monte -CarloVirtual

Irradiation

On-Orbit Error Rates(Creme -MC)

Mixed -SignalFunctional Rad

Models

TCAD-DrivenRad PDKModels

Failure Mechanisms

Rad-Aware VHDL

Radiation Aware Design

Simulation Enhanced Test

AvailableUnder developmentFuture research

A multi -agency-funded development effort is underway to integrate M&S into D&Q

A multi -agency-funded development effort is underway to integrate M&S into D&Q

Introduction to ISDE

Documents

Transcript of Introduction to ISDE