SAC Aging “2” ProjectSAC Aging “2” Project“Definition Stage” “Definition Stage”

Ready to go to Ready to go to “Implementation Stage”“Implementation Stage”

Joe Smetana

November 8, 2010

BackgroundBackground

• There is some evidence that the ATC performance of SAC solders tests is degraded by thermal preconditioning or aging– Multiple HDPUG Tests: SAC Acceleration Factors, Mild

Acceleration Test, and SAC Aging Test– CALCE Data (limited public)

• Auburn work using bulk solder samples shows a correlation between SAC solder properties and microstructures with aging. Alcatel-Lucent work using commercial components confirms microstructural changes with aging.

• Alcatel-Lucent (ALU) ATC testing shows that the magnitude of the aging effects are dependent strongly on the component type, CTE mismatch, nominal strain level, and ATC test parameters. The ALU work shows that in some cases, the effects due to in situ aging during thermal cycling control the failure process, not the effects due to the initial thermal aging, In some cases the ALU work showed aging effects with SnPb as well.

• The ALU work also indicates that it is necessary to track the microstructural evolution of the solder in order to understand the effects of thermal preconditioning.

BackgroundBackground

• The analysis of the current HDPUG SAC Aging Test is complicated by various factors including: – Early fails (left censored data) during non-monitored section of

ATC– Potential voiding issues– Very high strain components– Inadequate microstructural characterization and failure analysis.

The existing aging correlations are based strictly on statistical analysis of ATC data and are not supported by failure analysis or microstructural correlations.

The bottom line is that this has taken very long (project initiated ~2 years ago) and it is yielding ambiguous data. At best, this test produced only qualitative comparisons that will require additional follow-up testing.

Project OverviewProject Overview

• Use an existing test board design and component(s) to facilitate the launch of a new SAC Aging Project. The component(s) will be selected from ones that have shown signs of being impacted by aging in the Mild Acceleration Test– The test board design already exists and the components can be

procured easily. – The exact test board is completely compatible with the ALU ATC

chamber and rack system because it has already been cycled in previous test programs.

– The ATC program should be low cost and relatively easy to execute.

– This testing can address some of the existing gaps in the current SAC Aging project

• Perform baseline microstructural characterization on ambient (no age) and aged samples. Perform failure mode analysis and characterization on ATC samples to determine extent of microstructural evolution and impact on final failure.

• Build additional samples (5 boards for each component type) for microstructural analysis at different aging times/temperatures – complete plan to be worked out.

Test Vehicle (AT-1 Board)Test Vehicle (AT-1 Board)

• 16 each of 2 component types per board

• Test Board Size: 6.5” x 7”x .093” thick

• Connector not populated – for wiring only

• Extra ground lug point

Only 2 boards required for sample size of 32 componentsPopulate only one component type at a time

Test Board StackupTest Board Stackup

• Most routing on Layer 2• Layers 3 and 4 are planes• Open areas on L2 and 5 are thieved

ComponentsComponents

• 0.8mm pitch 192 CABGA – Available from Practical Components (Amkor)– Can be procured with “large” die (special order code) - Die size 475x475 mils

– Ball size: 0.46mm– Need to buy a minimum of 240– SAC 305 solder balls

• Same component used on the VIP test and on the Mild Acceleration Test– ATC performance known to be affected by preconditioning/aging

• Also test the BGA84 – but:– Will do this on a separate board set – in other words, we will build ½ of the

boards with BGA 192 components (only) and ½ of the boards with BGA 84 components only

• Fails much later than the BGA192 and we want to pull boards for FA without “excessive” extra cycles on them

• Will probably not increase the bare board costs (since the number of additional boards will likely still fall into a single lot build) – we only need a total of 30 – 15 for each component build.

• No extra component cost (we already have 284 of these)• Test would ATC longer to get failures on the BGA84.• Doubles slots in the ATC test chamber, Doubles wiring (but the total numbers are

still rather low)

DOE/Test OverviewDOE/Test Overview

Similar to SAC Aging 1 – but eliminates Alloy and Strain as Variables, uses only a single elevated temperature, two different ATC dwells, different component

ATC Profile: 0-100ºC, 10 and 60 minute dwells

2 boards required per test leg32 components per test leg

Factor A B C

Row #Time (Hrs)

Temp (°C)

Dwell (Mins)

1 480 25 10

2 480 25 60

3 240 125 10

4 240 125 60

5 480 125 10

6 480 125 60

AssemblyAssembly

• Assemble 30 total boards, 15 each with only the BGA 192 component, 15 each with only the BGA 84 component– 12 boards for testing + 2 boards for

profiling + 1 board for witness samples at each aging condition (3 x 2 component types) + boards (10) for aging/microstructure analysis at different times/temperatures

– Flextronics is already tooled for this assembly –

– Solder Paste – SAC 305

Microstructure AnalysisMicrostructure Analysis

• A key part of this is microstructure analysis. We will build 5 extra each board type for aging/pulling at different ATC cycles and/or different preconditioning (for example – 75C for TBD time to match similar microstructures seen etc.)

Cutting the board for Long Term Aging Cutting the board for Long Term Aging Samples (non-monitored)Samples (non-monitored)

Allows for 4 different “pulls” per board – total of 20 options with 50 boards

Cut lines

Draft Pull Schedule for Long Term Aging – Draft Pull Schedule for Long Term Aging – subject to changesubject to change

Draft Aging/Evaluation Schedule

 5 total boards of each type for aging/cross-sectioning. No monitoring  

  4 "sections" per board (20 total)  

 4 components per section for each type (BGA 192 on one board type, BGA84 on the other board    

Aging ConditionNumber of

boards

Number of Sections (4 components

each)Pull

Interval First Pull 2nd Pull 3rd Pull 4th pull 5th pull 6th pull 7th pull

75ºC 1 46

months6

months12

months 18 months 24 months      

50ºC 1.75 76

months12

months18

months 24 months 30 months36

months42

months48

months125ºC, 240 hours, +

ATC 1 4500

cycles First FailFF+500

cycFF+1000

cycFF+1500

cyc      125ºC, 480 hours, +

ATC 1 4500

cycles First FailFF+500

cycFF+1000

cycFF+1500

cyc      None (Orig Witness

sample) 0.25 1 NA              

                     

ATC Samples - in the 10 minute or the 60 minute dwell chambers? Could

split them 50/50?                    

Project Key StepsProject Key Steps

• Project Lead (ALU)• Design TV – Complete (ALU)• Build/Purchase Bare Boards: Meadville• Purchase BGA192 Components: Oracle • Assembly : Flextronics SJ• Precondition Boards: ALU• Wiring and ATC (0-100ºC, - 2 dwells (ALU)• Weibull Analysis (ALU)• Long Term Aging for Microstructure Analysis

(ALU)• FA and Microstructure Analysis (ALU,

Celestica, IBM and others?)• Final Report (Team)

Schedule -DraftSchedule -Draft

Est completion Actual

Design TV Complete

Receive Components 12/2010

Receive Boards 12/2010

Assembly Complete 2/2011

Precondition Boards 3/2011

Begin LT aging 3/2011

Wire Boards 3/2011

ATC start 4/2011

ATC complete 10 min dwell 11/2011

ATC complete 60 min dwell 7/2012

Begin Microstructure Studies 4/2011

Complete LT Microstucture Eval 3/2015

Weibull Analysis Ongoing with ATC

Interim report 9/2012

Final Report 2015?

TEAMTEAM

• Alcatel-Lucent

• Celestica

• Oracle

• IBM

• Flextronics

• Others…