inemi Lead-Free Rework Optimization Project: Solder Joint Characterization and Reliability

Transcription

1 inemi Lead-Free Rework Optimization Project: Solder Joint Characterization and Reliability Project Co-Chairs: Jasbir Bath, INEMI Craig Hamilton, Celestica Holly Rubin, Alcatel-Lucent

2 Phase 3 Participants Celestica Creation Technologies Flextronics Plexus Sanmina-SCI Alcatel-Lucent Intel Cisco Nihon Superior Senju 1

3 Acknowledgments Debra Fleming Richard Popowich Richard Coyle Mark Voss Alan Donaldson Akira Kita Takatoshi Nishimura Juthathip Fangkangwanwong Keith Howell Denis Jean Dale Lee Leo Anderson Fred Motaghian Khalid Mahmood 2

4 Outline Purpose and Background Accelerated Thermal Cycling (ATC) Test Details ATC Results Solder Joint and Plated Through Hole (PTH) Copper Knee Thickness Characterization 3

5 Overall Project Purpose Evaluate and recommend best practices, rework equipment requirements, impact of adjacent component temperatures and procedures for best practice lead-free rework processing. 4

6 Background Reliability tests in the previous NEMI lead-free assembly and rework project indicated a need to initiate a follow on inemi Rework Optimization Project. Thermal fatigue life of the CBGA937 component was reduced after rework of adjacent ubgas on the iinemi Payette board, possibly due to secondary reflow of the CBGA joints Thermal cycling of reworked PTH components was not performed due to hole fill and copper dissolution issues Phases 1 and 2, and Part 1 of this presentation, included rework equipment and rework process development that: Assured CBGA joint secondary reflow did not occur Improved PTH hole fill 5

7 Phase 3 Purpose Validate the developed rework processes by comparing ATC performance of: As-assembled CBGA SJR (solder joint reliability) to that of CBGA adjacent to reworked components selected PTH component SJR before and after rework solder joint characterization before and after ATC 6

8 Phase 3: ATC Testing Experimental variables: Number of times a component is reworked (0, 1, 2) Rework Solder Alloy (SAC305, Sn-Cu-Ni) A total of six rework operations were performed on each board that was reworked. Components monitored during ATC: CBGA937 BGA604 socket DIMM278 PDIP16 One daisy chain per component; daughter cards used to complete daisy chains for the DIMM and BGA socket 7

9 Payette Board Phase 3* 1X rework SAC305 DIP16 1X rework Sn-Cu-Ni DIP16 1X 2X 2X rework SAC305 BGA256 2X 1X CBGA 1X rework Sn-Cu-Ni DIMM278 1X Sn-Cu-Ni 1X SAC 305 1X rework SAC305 BGA socket 2X rework Sn-Cu-Ni DIP16 reworked Not reworked *125mil[3.2mm] thick, OSP, 7 x 17 inch 8

10 Design of Experiments Component Ref Desig. 1 st Pass Assy Alloy Rework Alloy µbga 1 U40 SAC305 SAC305 2X # Rework As-Assembled Sample Size not monitored Reworked Sample Size not monitored in ATC in ATC CBGA 1 U27 SAC305 N/A N/A BGA Socket U28 SAC305 SAC305 1X DIMM 1 J7 SAC305 Sn-Cu-Ni 1X DIMM 2 J8 SAC305 N/A N/A PDIP 1 U1 SAC305 Sn-Cu-Ni 1X * PDIP 2 U2 SAC305 Sn-Cu-Ni 2X * PDIP 3 U18 SAC305 SAC305 1X *sample size reduced due to assembly defects 9

11 Thermal Cycle Profile Thermal cycling parameters were selected to maintain consistency with prior testing: 0 to 100C, 10 min ramp, 10 min dwell Monitored by event detector Total cycles: 8700 as-assembled, 8200 reworked 10

12 ATC Results CBGA: only component with verified solder joint failures BGA socket, DIMM & PDIP: some failure events detected Failures not due to solder joint failures DIMM and BGA socket failures attributed to daughter card misalignment PDIP failures could not be electrically verified; high temperature electrical measurements, xray, and cross-sections revealed no failures in the solder joints or copper barrels/traces 11

13 SOLDER JOINT CHARACTERIZATION 12

14 BGA Socket Characterization, T0 4.1 m 4.1 m 4.1 m 4.1 m 4.1 m 4.5 m As-assembled, no ATC 2.50µm.288µm 7.68µm 3.52µm Reworked, no ATC Well formed solder joints with minimal voiding No significant IMC growth after rework 13

15 BGA Socket Characterization, post ATC AE29 AE2 AE2 AE16 AE29 Board 48 (no rework) AE29 Board 18 (rework) AE2 AE18 AE29 Well formed solder joints with minimal voiding Fatigue crack initiation 14

16 CBGA Baseline Microstructural Analysis Intermetallic Evolution in Bulk Solder 0 ATC cycles 15

17 CBGA Thermal Exposure on Reworked PCBAs The CBGA was not reworked; CBGA joint temperatures did not approach the solder melting point (<217 C) during BGA rework CBGA experienced thermal exposure when the entire PCBA was baked for various times/temps during rework of other components PCBA bake times and impact of BGA rework are likely to be the dominating thermal rework effects. Total PCBA bake times during rework (including times for 2x rework) are estimated as ~1.5-4 hours at 150C ~24 hours at 125C ~15-60 minutes at 105C Component Bake Method Bake Temp (C) Min Bake Time (min) Max Bake Time (min) BGA Oven BGA Socket None NA NA NA NA DIMM 1 Oven PDIP 1 Oven PDIP 2 Oven PDIP 3 Oven # Bakes 16

18 CBGA Characterization, T0 Baseline Post rework Bd44 AH31 Bd1 AH31 Overall good solder joint formation with minimal solder process voiding Some solder standoff variations (for both baseline and post rework) Higher Cu 6 Sn 5 intermetallic compound (IMC) particle density (dark phase) in the reworked solder joint. 17

19 Interfacial IMCs Component CBGA: Baseline vs Reworked, T0 4.0um Baseline Assembly Post Rework Bd1 AH31 3.8um Bd44 AH31 4.4um Bd1 AH31 4.8um 3.8um PCB 4.0um Bd44 AH31 Bd44 D31 AH31 No significant IMC growth in the post rework sample No evidence of secondary reflow Bd1 D31 18

20 CBGA: Baseline vs. Reworked No ATC Bulk IMC Baseline Post rework Component side Bd44 AH31 Component side Bd1 AH31 Ag3Sn There appears to be coarsening (growth) of the Ag3Sn IMC particles (light phase) post rework. IMC particle growth (both Cu6Sn5 and Ag3Sn likely is the result of thermal exposure during rework processing. IMC particle growth is a common effect due to thermal pre-aging. 19

21 CBGA Conclusions: Baseline and "reworked", 0 ATC cycles (pg 1 of 2) Acceptable solder joint formation with minimal voiding and some variation in standoff No evidence of secondary reflow due to rework processing. Typical intermetallic formation and thickness at PCB and package sides of the solder joint: PCB: 2-6 m on both boards Package: variable thickness & similar on both boards 20

22 CBGA Conclusions: Baseline and "reworked", 0 ATC cycles (pg 2 of 2) The Cu 6 Sn 5 IMC particles are more evident post rework. IMC particle growth likely is the result of thermal exposure during rework processing. IMC particle growth is a common effect due to thermal pre-aging but the Cu IMC particles are not considered a factor in providing fatigue resistance. The post rework samples show evidence of Ag 3 Sn IMC particle coarsening (growth). This Ag 3 Sn particle coarsening is characteristic of the Ag 3 Sn particle coarsening reported in isothermal aging studies. Pre-aged microstructures often exhibit moderate losses in thermal fatigue reliability depending on the component strain and temperature cycling parameters. 21

23 Baseline and "Reworked" CBGA "Reworked" As-assembled There is moderate, 10% reduction in reliability in the CBGA with thermal exposure due to rework processing on other components. The reduction in reliability could be the result of thermal pre-aging. The Weibull slope is significantly lower for the rework test cell. This effect has been associated with thermal pre-aging in previous studies. 22

24 CBGA Microstructural Analysis Post ATC 23

25 Baseline Assembly, 2134 cycles Bd22 T31 Post rework, 2032 cycles Bd52 T31 Solder fatigue failure mode Package side only on baseline PCBA Both package and board side on reworked PCBA Bd52 T31 There is no significant difference in the resultant microstructures in baseline or post rework samples after ATC (~2100 cycles). 24

26 CBGA: Baseline vs Reworked ATC Failure Mode Baseline Assembly, 2134 cycles Post rework, 2032 cycles component side Bd52 AH31 Bd22 AH31 Bd52 AH31 Bd22 AH31 PCB side There is no significant difference in the resultant microstructures in the strain-localized regions adjacent to the fatigue cracked region between baseline and post rework assemblies. 25

27 CBGA Summary and Conclusions On the "reworked" samples, Cu6Sn5 IMC particles were more evident and there was Ag 3 Sn IMC particle coarsening (growth) and degradation of Ag 3 Sn IMC particle network along the primary Sn boundaries. A moderate, 10% reduction in CBGA fatigue reliability was observed in the "reworked" samples. The reduction in reliability could be due to the Ag 3 Sn particle coarsening caused by thermal pre-aging from the thermal exposure during rework processing. Pre-aged microstructures often exhibit moderate loses in thermal fatigue reliability depending on the component strain and temperature cycling parameters. There is no significant difference in the resultant microstructures in baseline or "reworked" samples after ATC (~2100 cycles). This result is not unusual in temperature cycling comparisons of SMT and pre-aged microstructures. 26

28 PDIP Solder Joint Characterization 27

29 PDIP Cross Section Locations 1 8 Pin 1 Pin8 28

30 PDIP Time 0 SAC 305 SAC, initial wave SAC 1X rework Some shrinkage cracks with SAC Good hole fill, but significant lack of wetting in reworked samples at top of pin, due to flux application only to bottom side of board, but not to pins Increased voiding in reworked sample, typical of flux entrapment of reworked PTH components 29

31 PDIP Time 0 Sn-Cu-Ni 1X Sn-Cu-Ni 2X Sn-Cu-Ni Similar hole fill, wetting, and voiding for 1x and 2x rework When compared to initial SAC Assembly: Good hole fill but reduced wetting at top of lead due to flux application only to bottom side of board, but not to pins Increased voids, similar to 1X SAC rework samples 30

32 PDIP Post ATC As-assembled Post 8700 cycles U18 (SAC 305) Reworked Post 8200 cycles 2X Sn-Cu-Ni Reworked Post 8200 cycles 1X SAC 305 Similar fatigue cracking in as-assembled and reworked PTH joints after ATC 31

33 U2 (2X Sn-Cu-Ni) post ATC As Assembled After ATC cycles Reworked After ATC cycles Solder joint cracking similar as-assembled and reworked after ATC 32

34 Cu thickness measurement locations Top knee Top barrel Left Side Bottom barrel Right Side Bottom knee 33

35 A Comparison Of Reworked PDIP Cross-sections (Bottom Knee Measurements) (SAC 305 1X) (SAC 305 1X, Sn-Cu-Ni 1X & 2X) Unassembled (bare) board Cu thickness used as "estimator" for initial thickness The target minimum Cu thickness after assembly/rework is 0.5 mil For this PCBA and set of process conditions, the target is not met after 1X SAC305 rework process 34

36 A Comparison Of Reworked PDIP Cross-sections (Bottom Knee Measurements) (SAC 305 1X) (SAC 305 1X, Sn-Cu-Ni 1X & 2X) Unassembled (bare) board Cu thickness used as "estimator" for initial thickness The target minimum Cu thickness after assembly/rework is 0.5 mil For this PCBA and set of process conditions, the target is not met after 1X SAC305 rework process 35

37 A Comparison Of Reworked PDIP Cross-sections (Bottom Knee Measurements) (SAC 1X, 78sec) (Sn-Cu-Ni 1X, 30sec) (Sn-Cu-Ni 2X, 43sec) (SAC305 1X, 47sec) 78 sec SAC 1X rework leads to insufficient Cu knee thickness 47 sec SAC 1X rework still has risk of insufficient knee thickness Similar and sufficient Cu knee thicknesses for 1X and 2X Sn-Cu-Ni rework 36

38 DIMM Solder Joint Characterization 37

39 DIMM Cross Sections J J i S2 S1 Cross-section map X=no pin 38

40 DIMM Time 0 Pin#5 J8 S2 DIMM Time 0 J8, no rework (Same as J7) Representative images J8, adjacent to reworked J7 Pin#6 Representative J8 S1 images J7 reworked (1XSn-Cu-Ni) No obvious differences between J7/J8 as-assembled and J8 adjacent to reworked J7 Excellent hole fill and wetting Range of voids, constitute significant percentage of volume on "narrow" side of hole After rework: Range of wetting, from ~ ½ way up one side of the lead to almost complete wetting; due to flux application only on bottom side of board Pin#6 Pin#7 39

41 DIMM Post ATC as-assembled Bd 48 J8 S2 Pin 10 adjacent to reworked Bd 18 J8 S2 Pin 10 Reworked 1X Sn-Cu-Ni Bd 18 J7 S2 Pin 10 Similar fatigue behavior of as-assembled and reworked joints 40

42 Cu thickness measurement locations Top knee Top barrel Left Side Bottom barrel Right Side Bottom knee 41

43 A Comparison Of All The DIMM Cross-sections (Bottom Knee Measurements) (Sn-Cu-Ni 1X, 35sec) (Sn-Cu-Ni 1X, 25sec) Bare board again used as estimator for initial Cu thickness Expected loss of Cu after wave soldering Little additional loss of Cu after 1X Sn-Cu-Ni rework with a 25sec total contact time For this PCBA and set of process conditions, the total contact time must be less than 35 sec to achieve bottom knee thickness > 0.5 mil 42

44 A Comparison Of All The DIMM Cross-sections (Bottom Knee Measurements) (Sn-Cu-Ni 1X, 35sec) (Sn-Cu-Ni 1X, 25sec) Bare board again used as estimator for initial Cu thickness Expected loss of Cu after wave soldering Little additional loss of Cu after 1X Sn-Cu-Ni rework with a 25sec total contact time For this PCBA and set of process conditions, the total contact time must be less than 35 sec to achieve bottom knee thickness > 0.5 mil 43

45 A Comparison Of Reworked DIMM Cross-sections (Bottom Knee Measurements) (Sn-Cu-Ni 1X, 35sec) (SAC305, Not Reworked) (Sn-Cu-Ni 1X, 25sec) (SAC305, Not Reworked) For this PCBA and set of process conditions For both contact times, no significant differences between Cu thicknesses on reworked J7 and adjacent (not reworked) J8 At 35 sec contact time, minimum thickness requirement no longer met AND there is Cu knee dissolution on the adjacent DIMM (J8 ) 44

46 DIMM Post ATC Summary pg 1 of 2 For both as-assembled and reworked DIMMs: Post cycling, no cracks extend entire length or width of joint. Cracks can extend at least ¼ of the hole length from both the top & bottom. No observed cracks in Cu (traces or PTH barrel), either before or after ATC Adequate hole fill observed on cycled boards More extensive solder fatigue cracking is generally observed at top of PTH joint (package side), although some cracking is observed at PTH bottom. Tight process control required to ensure bottom Cu knee thickness remains above recommended 0.5 mil. 45

47 DIMM Post ATC Summary pg 2 of 2 More voids observed in J8 (as-assembled & adjacent to reworked J7) than in reworked J7; they constitute significant percentage of volume on "narrow" side of hole, but they had no observed impact on ATC results After rework: Range of wetting, from ~ ½ way up one side of the lead to almost complete wetting; possibly due to flux application only on bottom side of board; no impact on ATC results No significant differences observed in post cycled as-assembled and reworked cross-sections 46

48 Conclusions: SJ Reliability/Characterization) Rework had no observable impact on the solder joint reliability of the BGA socket or the PTH components under the applied test conditions The observed degree of wetting and voiding in reworked PTH joints did not lead to solder joint failures during the test time of this experiment A moderate (10%) reduction in CBGA solder joint fatigue reliability was observed in the reworked assemblies This may be attributable to microstructural changes within the joints caused by cumulative thermal exposure experienced by the CBGA during rework of other components Sn-Cu-Ni reduces Cu dissolution Tight process control is required to achieve bottom knee thicknesses > 0.5 mil on assemblies using thick PCBs and current PCB Cu thickness specifications. 47

49 Next Steps Continue to interact with IPC 6012 standard committee: 6012 may need to be updated to increase the minimum required Cu thickness on thick PCBs. This has been proposed to ensure sufficient copper knee barrel thicknesses after the wave soldering and wave rework soldering processes on large thermal mass network board type product whilst still providing a reliable solder joint. 48

50 contacts: David Godlewski Grace O'Mallley 49

51 Appendix 50

52 Component Details Component Ref.Des. Package Pitch Termination # I/O Size (mm) (mm) Finish/Ball Alloy Mfr µbga U40 17x SAC305 Topline or Practical CBGA U x SAC387 IBM BGA Socket U28 54x SAC305 Tyco DIMM J7, J x7.4 matte tin (over Ni) Tyco PDIP U1, U2, matte Sn (over 32x U18 Ni?) ChipPac 51

53 ATC Summary Results Component 1 st Pass Reworked Total Tested # Failures Verified Failures Total Tested # Failures Verified Failures CBGA BGA Socket DIMM PDIP

54 Reworked DIMM & PDIP Bottom Knee Cross-Sections (Sn-Cu-Ni 1X, 35sec) (Sn-Cu-Ni 1X, 25sec) (SAC 305 1X, 78sec) (Sn-Cu-Ni 1X, 30sec) (Sn-Cu-Ni 2X, 43sec) (SAC 305 1X, 47sec) 53

55 PDIP Cross-sections From Board 18 (Rework, Post /100 TC) (Sn-Cu-Ni 1X, 30sec) (SAC 305 1X, 47sec) (Sn-Cu-Ni 2X, 43sec) 54

56 A Comparison Of All The DIMM Cross-sections (Top & Bottom Knee Measurements) (Sn-Cu-Ni 1X, 35sec) (Sn-Cu-Ni 1X, 25sec) 55

57 DIMM Cross-sections From Board 1 (Rework, No TC) (SnCuNi 1X, 35sec) (SAC305, Not Reworked) 56

58 Cumulative Contact Time (sec.) Contact Time (sec.) Total Contact Times for DIP & DIMM Rework Boxplot of Contact Time - PDIP Boxplot of Contact Time - DIMM X Alloy X SN100C U1 (1X) SN100C U2 (1X) X SN100C U2 (2X) X SAC305 U18 (1X) X X 'X' = process time for cross-sectioned component(s) Green Bottom knee thickness > 0.5mil Yellow Bottom knee thickness Red Tail of distribution has knee thickness < 0.5 mil SN100C Alloy 57

59 contacts: David Godlewski Grace O'Mallley