Bad Schandau, Germany, May 30 – June 1, 2016 Page 2 Content Page 2 Background 1 Motivation 2 Experimental Approach 3 SRAM read disturb
Previous PDF | Next PDF |
[PDF] Ausgabe 9 September 2016 - Bad Schandau
9 sept 2016 · Nr 18/2016 2 Amtsblatt Bad Schandau Information i Aus dem Inhalt / Öffnungszeiten / Informationen aus dem Rathaus / Aus den
[PDF] Ausgabe 15 Januar 2016 - Bad Schandau
8 fév 2016 · 39 b Dienstag, den 19 01 2016, 16:00 - 18:00 Uhr Page 6 Nr 1/2016 6 Amtsblatt Bad Schandau Beschluss-Nr : 20151208 104 Beschluss –
[PDF] program 2016 - EUROPARC Federation
Page 1 1 TransParcNet Meeting 2016 Program Tuesday, 7 June commented walk with the mayor of Bad Schandau through the town 18:00 Opening at the
[PDF] Download Hausprospekt - Hotel Lindenhof Bad Schandau
ihr wohlverdienter Urlaub ist zu kostbar, um darauf zu verzichten Willkommen bei Freunden Page 4 Genießen Sie die Romantik
[PDF] AMTSBLATT - Bad Schandau
23 sept 2016 · Nr 19/2016 2 Amtsblatt Bad Schandau Information i Aus dem Inhalt / Öffnungszeiten / Informationen aus dem Rathaus / Aus den
[PDF] AMTSBLATT - Bad Schandau
5 avr 2016 · am 19 03 2016, 10:00 - 17:00 Uhr im NationalparkZentrum Sächsische Schweiz Dresdner Straße 2b • 01814 Bad Schandau Eintritt frei
[PDF] Global Thermal Think Tank, Germany 2016 - Global Wellness Institute
13 oct 2016 · She is the CEO of Toskana Therme Bad Sulza, Bad Schandau and Bad Orb, with its unique Liquid Sound Together with Micky Remann she
[PDF] Vielen Dank für Ihr Interesse AK Touristik ist seit mehr als 65 Jahren
Bad Schandau sagenhaftes Elbsandsteingebirge Diese romantische Landschaft, südlich von Dresden und an der Grenze zu Tschechien, zieht Jeden in
[PDF] 14th International Conference - IRSP 2016
Bad Schandau, Germany, May 30 – June 1, 2016 Page 2 Content Page 2 Background 1 Motivation 2 Experimental Approach 3 SRAM read disturb
[PDF] Bad Schwartau ist keine Marmelade
[PDF] Bad Soden 16.00 Hl. Messe in poln. Sprache 17.30
[PDF] Bad Soden mit Rhein- und Ahrtal Barockstadt Fulda
[PDF] Bad System config info - mega
[PDF] BAD THINGS ~ Bill Goodlad - Anciens Et Réunions
[PDF] Bad Tölz, 15.06.2017 Wissenschaftstage vom 21.06. bis 24.06. zum
[PDF] Bad Urach | Metzingen
[PDF] Bad Vilbel und Karben setzen auf Zusammenarbeit
[PDF] Bad Waldsee
[PDF] BAD WALDSEE / AULENDORF
[PDF] Bad Wildbad im Enztal
[PDF] Bad Wildungen: PreMedical-Active - gesund und fit mit
[PDF] BADABOUM : Fiche Technique - France
[PDF] BADDA-BOOM! BADDA - Anciens Et Réunions
Andreas Aal
1 , Gottfried Kurz 2 , André Clausner 3 1Electronic Analysis / Robustness (EEIP/1), Volkswagen AG, Berliner-Ring 2, 38436 Wolfsburg, Germany 2
GLOBALFOUNDRIES Dresden Module One LLC & Co. KG, Wilschdorfer Landstrasse 101, 01109 Dresden, Germany
3Fraunhofer Institute for Ceramic Technologies and Systems, Maria-Reiche-Strasse 2, 01109 Dresden, Germany
Phone: +49-(0)5361-9-38277; Fax: +49-(0)5361-9-57-38277; e-mail: andreas.aal@volkswagen.de Intermittent Functional Loss of Non-Degraded AdvancedSemiconductor Technology based Products
in Harsh Environments - Causes, Reproducibility, Mitigation 14 th International ConferenceReliability and Stress
-Related Phenomena inNanoelectronics
- Experiment and SimulationBad Schandau, Germany, May 30 - June 1, 2016
Content
Page 2
Background
1Motivation
2Experimental Approach
3 SRAM read disturb sensitivity under mechanical load 4 Primitive device electrical aging behavior under mechanical load 5Discussion
6Summary
7Page 3
1. Background
Future Mobility: Zero Emission, intuitiv, online
The future is a 4wheel computer with client & server functionality - the ultimate mobile devicePage 4
1. Background
Semiconductor Technology - Enabler for Functions that matter Automotive needs leading-edge technologies from a functional perspective, but ...Application CPU:
14/16 nm FinFET
(Samsung/TSMC)Baseband CPU:
20 nm SOC
Power Amplifier Modul:
100 nm GaN
LTE-Modul:
28 nm CMOS
NAND Flash:
15 -20 nm MLC Key -Technologies (i.e. Smartphone) Automotive: New digital Products & ServicesBig Data
Connected Car
Cloud- Computing
Autonomous
Driving
Page 5
2. Motivation
The challenge - AST* in automotive environments
This work: focuses on mech. induced parametric deviations ... ... function is bound to technology, but technology is bound to initial key-product design Reliability & performance of those technologies @ risk under automotive loadsRisks:
Stress exceeds strength
EOL reached too early (mission profile)
General mech. construction insufficience (cracks, delamination etc.)Parametric deviations
Permanent w/o aging effect
Reversible / intermittent
Partially permanent / reversible with aging effect *AST = Advanced Semiconductor TechnologiesPage 6
2. Motivation
The challenge - AST in automotive environments
Task: Mimic automotive loads and follow failure RCA* *RCA = Root Cause AnalysisApplication
Awareness of technology sensitivity to thermo-
mechanical stress insufficient AEC-Q100 qualification in sockets insufficient to mimic board -level effects under real reflow conditions Wafer technology qualification did not sufficiently consider CPI / CPBI* GapsSimilar effect observed
for 3 IC vendors / technologies in 2014 so not a single case * CPI / CPBI- chip package-board interaction Parametric drifts outside specified values after 1 st operationLucero, IRPS 2015
Leatherman, IRPS 2012
Page 7
3. Experimental Approach
Quantitative mech. loads & typ. environmental loads Quantify stress, watch effects - conclude knowledge basedApproach I
Analysis of SRAM read disturb sensitivity under mechanical loadNano-indentation
VdipR - Tests
FEM Simulation of mechanical induced stress @ transistor level caused by external forces Calibration of simulation & electrical measurementsApproach II
Analysis of primitive device electrical aging behavior under mechanical load Effect of uHAST, TC, wafer thinning on HCI behaviorPage 8
4. Experimental - SRAM read disturb
Approach I - mechanical setup for n-indentation
Application of mech. stress by n-indentation @ chip back side28 nm HKMG 64 Mbit SRAM, full process flow
Wafer thinning to 250 µm / Flipchip assembly to 948µPGA packageATE test at 85°C / 25°C of assembled SRAMs
Remove of package lid and further down thinning of remaining Si- thickness (min. 35 Ɋm)RD - Procedure (RDP)
Write step @ V
nom (checkerboard pattern)Voltage dip down to VdipR
Read (disturb) step
Voltage rise to V
nom & read with pass/fail assessmentRepeat @ different mech. load conditions
Calibration I
VdipR chosen at threshold to bit flip (high sensitivity to mech. load) Drawback - background noise solution: statistical averaging by repeating RDP (fail assessment when > 10 fails in 20 repetitions or # fails per 50VdipR repetitions )
Page 9
4. Experimental - SRAM read disturb
Approach I - VdipR Test
Fail calibration to mech. load
Page 12
4. Results
- SRAM read disturb Approach I - Simulation of mechanical induced stress @ transistor level caused by external forcesImage overlay of
VdipR and von-Mises stress simulation @ 1.3 N
Page 13
4. Results
- SRAM read disturb Approach I - Simulation of mechanical induced stress @ transistor level caused by external forcesImage overlay of
VdipR and hydrostatic stress simulation @ 1.3 N
Page 14
4. Results
- SRAM read disturb Approach I - Simulation of mechanical induced stress @ transistor level caused by external forcesImage overlay of
VdipR and normal stress in x direction simulation @ 1.3 NPage 15
4. Results
- SRAM read disturb Approach I - Simulation of mechanical induced stress @ transistor level caused by external forcesImage overlay of
VdipR and normal stress in y direction simulation @ 1.3 NPage 16
4. Results
- SRAM read disturb Approach I - Simulation of mechanical induced stress @ transistor level caused by external forcesImage overlay of
VdipR and normal stress in z (indentation) direction simulation @ 1.3 NPage 17
4. Results
- SRAM read disturb Approach I - Calibration of simulation & electrical measurements Correlation of the SRAM functionality & simulation stresses through chip operation voltage shift with/without indenter load Determination of voltage shift required to hold cell stable (increase under indentation load) Shift in minimum required SRAM operation voltage with simulation stresses reliability criteria can be derivedPage 20
5. Results
- TQV test structure chipApproach II - pure HCI aging, higher stress
HCI effect per channel length with distinct L - separationWatch these curves on the next slide
Page 21
5. Results
- TQV test structure chip Approach II - HCI aging after uHAST, higher stressTC/uHAST cause a ~const. shift of pMOS I
d,sat degradationThis form of graphical illustration is
insufficient to show the important thingsThese curves have shifted upwards
Page 22
5. Results
- TQV test structure chipApproach II - pure HCI aging
Effect of HCI stress on pMOS I
d,sat degradationPage 23
5. Results
- TQV test structure chipApproach II - HCI aging after uHAST / TC + uHAST
Partially reversible shifts and variance increase
Page 24
5. Results
- TQV test structure chipApproach II - HCI aging after uHAST / TC + uHAST
Shift & variation are voltage dependent
Page 25
5. Results
- TQV test structure chip Approach II - HCI aging & uHAST, TC, wafer thinningSecondary stress effect is L and V dependent
Page 26
6. Discussion
Approach II
Usually package form-factor related tests are applied after tech-qual rather than investigating the effect already during technology qualification This approach then considers mechanical effects as linear, reversible and without effect on aging In addition the meaning of variability increase / decrease my be underestimated TC after wafer processing before wafer thinning & further assembly may reduce variabilityResults from Intel (IRPS 2012) show that nMOS
I dsat shift is stronger affected as pMOS - we see pMOS sufficiently enough affected Shift is 37-50 % higher when going down from 25 °C to -10 °C Shift is 20 % higher when die thickness is reduced from 200 um to 120 umPage 27
7. Summary
Approach I
Based on the combination of a smart n-indentation setup, optimized SRAM sensitivity, FEM simulations and corresponding calibration via cell operation voltage adjustment, ...... it is now possible to quantify the transfer ratio of externally applied stress to local stress on Si device level, which ...
... can positively extend current DfR methodologies