IEEE ECTC June 4, 2004, Las Vegas, Nevada

Transcription

1 Reliability Testing and Data Analysis of an 657CCGA (Ceramic Column Grid Array) Package with Lead-Free Solder Paste on Lead-Free PCBs (Printed Circuit Boards) IEEE ECTC June 4, 004, Las Vegas, Nevada

2 Study Objectives To perform temperature cycling test and statistical analysis of an 657CCGA package on PCBs with HASL(Hot-Air Solder Leveling)-SnCu, NiAu (Electroless Ni and Immersion Au), and OSP (Organic Solderability Preservative) finishes. For a given confidence level, determine the population mean, true characteristic life, and true Weibull slope of the 657CCGA solder joints. To determine the confidence for comparing the quality (mean life) of the 657CCGA solder joints on PCBs with different surface finishes.

3 Contents Introduction 657CCGA Components Test Board Test Chamber and Temperature Cycling Data Acquisition System and Failure Criterion Reliability of 657CCGA Solder Joints (A) Simple Statistics (B) Statistical Analysis of 657CCGA Solder Joints (C) Failure Modes Summary and Recommendation

4 IBM s High-Density High-Performance 657CCGA Package 4.5+/-0.5mm X.0mm.0mm.0 mm 40X.0mm Solder Columns - (0Sn90Pb) 657 Places.+ 0.5/-0.08mm

5 Test Board with 657-Pin Ceramic Column Grid Array (CCGA) Packages CCGA Solder Pastes: Sn37Pb Sn3.9Ag0.6Cu

6 PCB and 657CCGA Sample Size Table PCB and package sample size details Number of Boards Board Finishes Solder Paste Number of Components 5 OSP SnPb 5 OSP SnAgCu 5 ENIG SnAgCu 5 SnCu HASL SnAgCu

7 TC (Air) 4 DOOR TC TC TC 3 TC Test Boards in Thermal Cycling Chamber (topview) and Thermal Couples Locations In Air Location 3 On Board Location Location 4

8 657CCGA Test Results (0 to 00 o C, 40 min.) Table 657CCGA thermal cycling test results (0 to 00oC, 40 min.) Solder Paste SnAgCu SnPb SnCu HASL PCB ENIG (NiAu) PCB OSP (Entek) PCB NA NA

9 Reliability The reliability of solder joints of a particular package on a PCB is defined as the probability that the solder joints will perform their intended function for a specified period of time, under a given operating condition, without failure. Numerically, reliability is the percent of survivors, R(t)= F(t) R(t) is the reliability (survival) function F(t) is the cumulative distribution function (CDF), representing the percentage of failures.

10 Order Statistics (Ranking) ( z) n nz( z) n n( n ) ( n j + ) z ( j )! j n( n ) z! ( z) n j+ ( z) = G n j = the failure order number n = number of samples G = required ranking z = the percent rank of the jth value in n Median rank ~ (j - 0.3)/(n + 0.4)

11 657CCGA with SnAgCu Paste on HASL-SnCu PCB (90% Confidence) Percent Failed % Rank 90% Confidence Intervals 5% Rank β=6.7507, η=567.95, ρ= β=0.0634, η= , ρ= β3=5.68, η3= , ρ=0.989 John Lau Agilent Cycle-to-Failure

12 Mean Life, Characteristic Life, and Weibull Slope of the 657CCGA Solder Columns with SnAgCu and SnPb Solder Pastes on Various PCBs (90% Confidence Intervals) SnAgCu paste on HASL PCB SnAgCu paste on NiAu PCB SnAgCu paste on OSP PCB SnPb paste on OSP PCB Sample Mean Life (µ), cycle Percent failed at mean Population µ at 90% confidence µ µ µ µ 46 7 Characteristic Life (θ), Sample cycle True θ (cycle) at 90% confidence 59 µ µ µ µ Sample Weibull Slope (β) β error (%) at 90% confidence True β at 90% confidence β β β β 7.5

13 657CCGA with SnAgCu Paste on NiAu PCB (90% Confidence) Percent Failed % Rank 90% Confidence intervals 5% Rank β=5.3884, η= , ρ=0.839 β=8.8795, η= , ρ= β3=3.7353, η3= , ρ= John Lau Agilent Cycle-to-Failure

14 657CCGA with SnAgCu Paste on OSP PCB (90% Confidence) β=8.9868, η= , ρ= β=3.7, η=37.75, ρ= β3=6.8698, η3= , ρ= Percent Failed % Rank 90% Confidence Intervals 5% Rank 0.0 John Lau Agilent Cycle-to-Failure

15 Mean Life, Characteristic Life, and Weibull Slope of the 657CCGA Solder Columns with SnAgCu and SnPb Solder Pastes on Various PCBs (90% Confidence Intervals) SnAgCu paste on HASL PCB SnAgCu paste on NiAu PCB SnAgCu paste on OSP PCB SnPb paste on OSP PCB Sample Mean Life (µ), cycle Percent failed at mean Population µ at 90% confidence µ µ µ µ 46 7 Characteristic Life (θ), Sample cycle True θ (cycle) at 90% confidence 59 µ µ µ µ Sample Weibull Slope (β) β error (%) at 90% confidence True β at 90% confidence β β β β 7.5

16 Weibull Slope Error (E) E N π e t dt = ( + C) π t E N e dt = Z( x)( bt + bt + b3t + b4t + b5t Z( x) = e π x p =.3649 b = b = b 3 = b 4 = b 5 = ε(x) < 7.5x0-8 t = + px ) + ε ( x) N = number of failures C = the required confidence level

17 657CCGA with SnAgCu Paste on HASL, NiAu, and OSP PCBs (Median Rank) OSP NiAu Percent Failed HASL β=6.7507, η=567.95, ρ= β=5.3884, η= , ρ=0.839 β3=8.9868, η3= , ρ= Cycle-to-Failure John Lau Agilent

18 Confidence Level (P) for Comparing the Quality (Mean Life) of Two Sets of Solder Joints P = + log log q q q = 5 t / 7 [ ρ ] + T t = T = ( r )( r ) ρω + Ω ρ = µ µ Ω Γ( + / β ) = Γ ( + / β) Ω Γ( + / β ) = Γ ( + / β ) µ and µ are the sample mean lives, β and β are the sample Weibull slopes, r and r are the number of failures, respectively, of sample and sample. T is called the total degrees of freedom.

19 Mean Life Ratio of the 657CCGA Columns with SnAgCu and SnPb Pastes on Various PCBs Different Assembly SnAgCu paste on HASL PCB SnAgCu paste on NiAu PCB SnAgCu paste on OSP PCB SnPb paste on OSP PCB Assigned Symbol Mean Life, cycles Mean Life Ratio (Compariso n) Determined Confidence A 397 B/A =.33 99% B 379 B/C =.4 89% C 793 C/A =.7 94% D 3760 D/C =.35 99%

20 657CCGA with SnPb Paste on OSP PCB (90%Confidence) Percent Failed % Rank 90% Confidence Intervals 5% Rank β=5.98, η= , ρ= β=8.566, η= , ρ=0.954 β3=4.0308, η3= , ρ= Cycle-to-Failure John Lau Agilent

21 SnAaste SnPb aste 657CCGA with SnAgCu and SnPb Pastes on OSP PCB (Median Rank) SnPb SnAgCu Percent Failed β=5.98, η= , ρ= β=8.9868, η= , ρ= Cycle-to-Failure John Lau Agilent

22 Mean Life Ratio of the 657CCGA Columns with SnAgCu and SnPb Pastes on Various PCBs Different Assembly SnAgCu paste on HASL PCB SnAgCu paste on NiAu PCB SnAgCu paste on OSP PCB SnPb paste on OSP PCB Assigned Symbol Mean Life, cycles Mean Life Ratio (Compariso n) Determined Confidence A 397 B/A =.33 99% B 379 B/C =.4 89% C 793 C/A =.7 94% D 3760 D/C =.35 99%

23 Cross Section of the Failed Columns

24 CONCLUSIONS and RECOMMENDATIONS At 90% confidence (i.e., in 9 out of 0 cases), the true characteristic life of the CCGA solder columns with SnAgCu solder paste on HASL (SnCu) PCB is larger than 59 cycles and less than 786 cycles. The true mean life can be as low as 069 cycles and as high as 55 cycles. The true Weibull slope is larger than 4.5 and less than 9.5. Similarly, at 90% confidence, the true characteristic life of the CCGA solder columns with SnAgCu solder paste on NiAu PCB can be as low as 984 cycles and as high as 3673 cycles. The true mean life can be no less than 655 cycles and no more than 3497 cycles. The true Weibull slope is larger than.86 and less than 7.9. At 90% confidence, the true characteristic life of the CCGA solder columns with SnAgCu solder paste on OSP (Entek) PCB is larger than 677 cycles and less than 37 cycles. The true mean life can be as low as 49 cycles and as high as 307 cycles. The true Weibull slope is larger than 5.66 and less than.3.

25 CONCLUSIONS and RECOMMENDATIONS At 90% confidence, the true characteristic life of the CCGA solder columns with SnPb solder paste on OSP (Entek) PCB is larger than 346 cycles and less than 4500 cycles. The true mean life can be as low as 307 cycles and as high as 467 cycles. The true Weibull slope is larger than 3.33 and less than 7.5. The quality (mean life) of the CCGA solder columns (with SnAgCu solder paste) on the NiAu PCB is better than that on the OSP with 89% confidence and that on the SnCu HASL with 99% confidence. In 94 out of 00 cases, the quality of the CCGA solder columns (with SnAgCu solder paste) on the OSP PCB is better than that on the SnCu HASL PCB. The quality of the CCGA solder columns with SnPb solder paste on OSP PCB is better than those with SnAgCu solder paste. This agreed very well with finite element simulation results reported elsewhere. More thermal cycling tests and finite element simulations should be done on lead-free assemblies with different dwell times.