The new JEDEC JESD89A Test Standard: How is it ...

The new JEDEC JESD89A Test Standard: How is it different than the old one and why should we use it? Robert Baumann Texas Instruments, Dallas Texas Dedicated to the memory of Mike Maher 1

RADECS 2007, Deauville

In Memory of Mike Maher

2


Outline • Motivation – do we really need it? • Review of the basic Physics & Test Methods • History/Overview – JEDEC JESD89 (released 2001) – JESD89A (released 2006)

• Future Challenges – JESD89B????

3


Motivation for JESD89 • SER produces the highest failure rate in commercial electronics of all other reliability mechanisms combined. • As SER became more important, companies that knew how to do the tests were at a disadvantage since companies that did no or incomplete testing reported, erroneously, very low SER – leading to customer confusion. • Usually when accelerated and even system tests are done improperly the result is an underestimate of the actual product failure rate. • Military and Aerospace had some well established standards and it was time for the commercial world to have a standard for the terrestrial environment 4


Example of Misinformation A certain DRAM vendor claimed: Based on ASER with alpha particles only < 50 FIT product SER SER of 50 FIT @ 90% CL based on 0 fails @ 450.8 k dev-hrs. So let’s do the calculation (assuming 0 errors after 450.8k dev-hrs.):

χ2 2.31 FR 90 < = = 5124 Fits !!! 3 2 ⋅ ( dev − hrs.) 450.8 × 10 dev − hrs How long a test would we need to have 90% CL that we had a 50 FIT part?

χ2

2.31 Dev − hrs > = 2 ⋅ (failure rate ) 50 Fits =

2.31 5 × 10

−8

e / hr


~ 47 Million dev − hrs .

That’s 1000 units for more than 5 years!!!!

5

Historical Example Nuclear physics in a Business Magazine!!!! Sun Screen Daniel Lyons, Forbes Global, 11.13.00 mysterious glitch has been popping up since late last year...web companies, telecommunications companies, a Baby Bell in Atlanta, an Internet domain registry on the East Coast, high-end servers made by Sun Microsystems have, for no apparent reason, suddenly crashed...It has caused problems for America Online, Ebay and dozens of other major corporate accounts…The Sun has caused crashes at dozens of customer sites. An odd problem involving stray cosmic rays and memory chips in the flagship Enterprise server line…Sun found it had been shipping servers whose cache modules contained faulty SRAM (static random access memory) chips from a supplier it won't name.

Loss of customer confidence

=

Loss of revenue 6


The Terrestrial Environment and Basic Testing Methods

7


The 3 Radiation Mechanisms Alpha Particles

10B

• Directly ionizing • Emitted from U,Th impurities

He

and Low Energy Neutrons

• σth 10B is huge & highly ionizing emissions • High 10B concentration in BPSG (4 - 7%)

• Peak stopping pwr ~ 16 fC/um

• Peak stopping power ~ 16 & 25 fC/um

• limited range (< 40 um)

• Effect localized (< 5 um)

• Dominant if not screening for alphas

• Dominant in parts using BPSG

Li

n 28Si,

O, Cu, W, etc.

n

High Energy Neutrons • Complex reactions 10B

• Secondary products cause SEE • Stopping power > 100 fC/um • Effect increases with altitude • Cannot easily be shielded RADECS 2007, Deauville

He 8

Real-time (System) Testing Cosmic background

Hundreds or thousands of units for hundreds or thousands of hours!!!

SEU/MBU/SEL rate under actual use conditions 9


Monoenergetic (Proton) testing Intense Monoenergetic Radiation Source

Device cross-section

∞

∫ σ ( E ) ⋅ F ( E )dE Ec RADECS 2007, Deauville

SEU/MBU/SELU rate under actual use conditions One test condition = one data point

10

Broad Spectrum (neutron) Test Intense “White” Radiation Source Neutrons/cm2-sec-MeV

1E-02

1 multiplication

1E-03

(Acceleration Factor)

1E-04

SEU/MBU/SELU rate under actual use conditions

1E-05 WNR Beam / 1.38E08 Atmosphere

1E-06 1E-07 1

10

100

Neutron Energy (MeV) RADECS 2007, Deauville

1000 11

JEDEC Test Spec. JESD89 History, Revisions, & Future

12


Goal of JEDEC JESD89

“Radiation Test specification? Just define a radiation exposure and measure the response… Right?” 13


Goal of JEDEC JESD89 “Making a Soufflé au fromage? Just eggs and cheese…right?”

14


Goal of JEDEC JESD89 “Making a Soufflé au fromage? Just eggs and cheese…”

15


Goal of JEDEC JESD89 To develop a “cook book” that includes enough details to enable people, even non-experts, to make and report radiation characterizations and extrapolations that are accurate (to some level) and inclusive of all mechanisms.

16


Specific Goals • To provide guidance concerning types of radiation testing required, sample sizes, facilities, etc. • Ensure that alpha particles and neutron reactions be accounted for in separate measurements - not allowing total SER to be reported without both effects. • Reduce 100x - 1000x reporting errors to < 20x*. • Eliminate confusion - educate customers & vendors. * Experienced investigators should be able to get within 2x of the “real answer” but the 20x factor assumes a neophyte with incomplete assumptions; e.g. actual package alpha emission is often underestimated by 10x or more, so trying to keep the worst-case error to < 20x is reasonable. Likewise, since no guidance is given on how to deal with thermal neutron effects in the first JESD89, there is another 2x-8x factor that might occur.

17


JESD89 History • Started and initially led by NASA HQ in 1999 by Michelle Gates in an effort to get commercial companies to characterize terrestrial SER performance. Mike Maher provided strong guidance on JEDEC specific issues. • Under Semiconductor Research Committee (SRC). • Project was taken over by Baumann in early 2000. • Original JESD89 was released August 2001. • Took ~ 20 people nearly 3 years to finish. 18


Scientist vs. Engineer Fully understand all mechanisms… Absolute dosimetry… Absolute magnitudes !

Relative magnitude ok! Quick approximations… Need working “standard” model now!

19


JESD89 Sections Introduction, Introduction, definitions definitions

System System (Real-time) (Real-time) SER SER Alpha AlphaParticle ParticleSER SER

Appendices Appendices

Neutron Neutron(Proton) (Proton) SER SER

20


Big Impact of JESD89 • First comprehensive standard for SER in commercial parts in the terrestrial environment. Includes chapters on system testing (real-time), neutron testing, and alpha testing. • Standard environment defined (NYC @ sea-level) with instruction and tables to handle other geographical locations. Allowing “apples-to-apples” comparisons. • Specified that alpha and neutron effects MUST both be characterized before reporting a part’s SER. • Errors in SER estimates constrained to < 20x* • Key Complaints: too long, too hard to understand, not international enough. * Experienced users would be within 2x or better of the correct SER values. RADECS 2007, Deauville

21

JESD89A History • Started 2004 in an effort to address errors, improve models, and address some of the user concerns. • Effort took ~ 70 people 2 years to finish. An improvement over the original JESD89 process, especially considering the much larger working group. • Went through JEDEC process quickly due to politically savvy and strong support from Mike Maher and Charlie Slayman. • Released October 2006.

22


New in JESD89A • Almost complete rewrite (60 pages to 80 pages) • New chapter on experimental set-up • Updated chapter on system level testing • Abolished 3/3 minimum requirement for ASER (devices/lots) • Updated chapter on alpha particles with emphasis on geometry factor. • More accurate neutron model after Goldhagen • Neutron cut-off kept at 10MeV • Updated proton cross-section method - added requirement of using a 150MeV proton energy • New chapter on 10B thermal neutron reaction • Expanded facilities list on website. Website with easy “geographical factor”. ( www.seutest.com ) RADECS 2007, Deauville

23

JESD89A Sections Introduction, Introduction, defintions defintions

System System (Real-time) (Real-time) SER SER

Experimental/Package Experimental/Package

Alpha AlphaParticle ParticleSER SER

Neutron Neutron(Proton) (Proton) SER SER

Appendices Appendices

Thermal ThermalNeutron NeutronSER SER 24


What is NOT covered adequately in JESD89A – or, What we will have to do for JESD89B

25


Alpha vs. Neutron SER 5 Rel. SER (aSER/nSER)

ULA materials

4 3

65nm @ 1.1V

Alpha dominated SER

180nm @ 1.5V

2 1 0 0.0001

0.001 Alpha Emission (a/cm2-hr)


0.01

Neutron dominated SER 26

Error Due to Absorption Error Factor based on geometry and absorption vs. geometry only (JESD89A compliant) No chip layers Separation (mm) 0.01 0.1 1 2 4 6 8 10 12 14 16

8µm

8µm + 5µm polyimide

Geometry + Absorption

Geometry 1.01 1.05 1.53 2.18 4.14 6.67 10.32 14.18 20.53 28.30 34.48

4µm

1.04 1.11 1.45 1.56 1.55 1.62 1.64 1.71 1.63 1.60 1.71

1.53 1.67 2.08 2.19 2.06 2.23 2.19 2.36 2.39 2.34 2.69

2.34 2.55 3.07 3.20 3.03 3.18 3.24 3.56 3.71 total absorp. total absorp.

3.49 3.53 4.47 4.89 4.59 4.63 5.08 5.53 total absorp. total absorp. total absorp.

232

“Determination of Geometry and Absorption Effects and their Impact on the Accuracy of Alpha Particle Soft Error Rate Extrapolations”, R. Baumann and D. Radaelli, 2007 IEEE NSREC


5mm diameter Th alpha particle source 5 x 10 mm device with various chip layers 27

Neutrons Effects at < 10 MeV -2

Diffential flux (/(cm².MeV.s)

10

Flux Gordon'04 WNR spectrum (/3E8)

-3

10

-4

10

-5

10

-6

10

∫ ∫

1000

Flux = 13 n/cm²/h

10

-7

10

-8

10

-9

10

“Single Event Upsets Induced by 1-10 MeV Neutrons in Static-RAMs Using Mono-Energetic Neutron Sources” J. Baggio, D. Lambert, V. Ferlet-Cavrois, P. Paillet, C. Marcandella, O. Duhamel, 2007 IEEE NSREC

0

10

1

10

2

10

3

10

10

Flux = 7 n/cm²/h

1

Energy (MeV) • The contribution of 1-10 MeV neutrons to SER is no longer negligible: – Experiments & simulations show > 10 % effect • MC simulations show that (n, α) drives the cross-section in 4-6 MeV range RADECS 2007, Deauville

28

Shielding Effects for Neutrons Sensitivity to lower energy neutron events has other effects…

• 1998 Eight boards in neutron beam ~ 33% attenuation from 1st to last board. • 2006 After 1st board in neutron beam ~ 25% attenuation • Building effects more important?

29


Improvements for JESD89B • Better clarification on event types (SEL, SET, SEFI, etc.) • Update chapter on alpha particles for absorption effects. • Re-evaluate neutron cut-off of 10MeV…should we use lower cutoff? Work by Baggio et al. suggests we should. • Neutron shielding effect guidelines. • Improve chapter on 10B thermal neutron effects perhaps with guidence on dosimetry and absolute method. • “Gold chip” standard SRAM test chip? • Further expanded facilities list on website • Create standard reporting document example? • Correct errors & improve clarity… 30


Summary • JEDEC JESD89 (August 2001) was the first test spec. for the commercial industry in the terrestrial environment. Created by a relatively small group of predominantly U.S. people over 3 years. • It reduced SER miscalculations (observed to be as high as 1000x) to < 20x in most cases (for experienced users error is < 2x). But ignored some and effects and was not as international as it should have been.

• JESD89A (2006) involved a large ~ 70 person group of people from many countries. • It reduces SER miscalculations < 5 x (better than this for experienced users) and further standardizes SER characterization techniques.

• JESD89B: There is a need for further revisions in the future. We have pointed to some of the areas for improvement. 31 RADECS 2007, Deauville