1 Material Testing & Mitigation Techniques for Pad Crater Defects John McMahon P.Eng.
Agenda Introduction Pad Crater Defects Bend Test Methods Process Strain Limits Laminate Susceptibility Package Compliance Observations Mitigation Techniques 2
Pad Crater Defects Crack initiation in the PWB resin system between the top surface and the first layer of glass reinforcement Pb-free solders are less compliant IMC interfaces are now more resistant to fracture Packaging houses have converted to direct solder on Copper processes Phenolic cured resins systems replace Dicy cured resins Micro clays and ceramic particles are added to resin systems to reduce Z-axis expansion. Pad Crater is now reported as the dominant mechanically induced defect in PCBA assembly Courtesy of the Intel pad crater group 3
Bend Test Comparison 4 point Bend Unidirectional strain and constant bending moment between anvils Spherical bend Based on ring on ring stress test. Drives biaxial strain Proposed specification 9707 for spherical bend Spherical Bend Test Geometry 120 mm 4 SMTAI Oct. 2011
Crack Loading & Path Crack path is determined by two primary factors. Stress profile relative proportions of mode 1 & mode 2 Stiffer packages generate mode 1 dominant sytems. Type of material filled or unfilled illed material creates many opportunities for crack to turn Mode 1 Opening Mode 2 In-Plane Shear Three Modes of Crack Loading Pad crater Cohesive & Adhesive Mode 3 Out-of-Plane Shear Typical Telecomm product There is only 10 to 20 micron between Pad Peel & Pad Crater Pad Peel or Pad Lift Cohesive only 5 Characteristic crack paths
Strain Limit Specifications IPC-9704 process window for strain vs. strain rate based on failure in solder joint Revised guidelines published in 2005 Courtesy of Keith Newman Current proposed revision of IPC 9704 moves strain vs. strain graph to white paper to make updating easier 6
Celestica Internal Strain Limits Test Program to Validate Safe Working Strain Limits for Pb-free compatible materials. Spherical bend Test geometry Step stress approach to testing stiff systems. Test program: 3 strain rates 1000, 3000, 6000 2 board thicknesses 0.100 & 0.130 2 surface finishes 4 solder sphere alloys Primary attach & forced rework 1000 Strain (ue) Typical curves for a single board thickness SnPb Guideline Below 2000 ue/s the results seem to be in good agreement with previous guidelines. Above 2000 ue/s The difference becomes significant. 500 "Diag 0.1% 90%CL "P1 0.1% 90%CL" "Survival" "ailure" Log Strain rate (ue/s) 0 100 10000 7
Laminate Susceptibility Multiple Suppliers build to a single design data package With minor exceptions the slopes (Beta) are very similar and we understand now what to expect from a new lot. The test method does define statistical differences between lots. ailure, (ue) Comparisons are based on: Nominal Principal Strain rate = 3000 µe/s TV - 20 layer 2.54 mm (0.100) (0.104) 32 mm CBGA Probability Weibull 2P Probability-W eibull ab 1\Lam A =13/S=27 ab 3\Lam D =14/S=14 ab 3\Lam G =19/S=9 ab 4\Lam C =12/S=20 ab 4\Lam H =10/S=14 ab 5\Lam B OSP =32/S=16 ab 6\Lam E =16/S=16 ab 6\Lam =11/S=17 micro strain, (ue) 8
Laminate Susceptibility Multiple Suppliers build to a single design data package This view is perhaps more easily read in terms 1.000 of risk / reliability There is some question about extrapolation to very low values when the slope is different. 36.000 29.000 22.000 15.000 8.000 9 Beta Confidence Contours Eta ab 1\Lam A =13/S=27 ab 3\Lam D =14/S=14 ab 3\Lam G =19/S=9 ab 4\Lam C =12/S=20 ab 4\Lam H =10/S=14 ab 5\Lam B OSP =32/S=16 ab 6\Lam E =16/S=16 ab 6\Lam =11/S=17 17/02/2011 2:06:55 PM 1.000 500.000 750.000 1000.000 1250.000 1500.000 1750.000 Comparisons are based on: Nominal Principal Strain rate = 3000 µe/s TV - 20 layer 2.37 mm (0.100) (0.104) 32 mm CBGA Reliability No Damage Plot micro strain, (ue) ab 1\Lam A =13/S=27 Reliability Line ab 3\Lam D =14/S=14 Reliability Line ab 3\Lam G =19/S=9 Reliability Line ab 4\Lam C =12/S=20 Reliability Line ab 4\Lam H =10/S=14 Reliability Line ab 5\Lam B OSP =32/S=16 Reliability Line ab 6\Lam E =16/S=16 Reliability Line ab 6\Lam =11/S=17
Laminate Results Estimated Principle Strain A B S There is no notable difference in the structures that would account for a difference of 100% between laminate B & C. The hidden factor is the resin system. 10 SMTAI Oct. 2011 D&P ailed Survived C S D S E S S G S H S S
Package Compliance Selected Laminates No change in rank order. More compliant pkg. produces an increase in survivable strain. In this case 14%-17% Eta. Exception is at the high end of the range. Distributions converge as mixed failure mode is more probable. 40.000 ailure, (ue) 99.000 90.000 50.000 Contour Monolithic Ceramic Built up organic w/ metal lid Probability-W eibull olio2\lam_g metal =17/S=7 olio2\lam_b metal =14/S=16 olio2\lam_e metal =16/S=8 olio2\lam_g cer =19/S=9 32.600 10.000 olio2\lam_g metal =17/S=7 olio2\lam_b metal =14/S=16 olio2\lam_e cer =20/S=16 Beta 25.200 17.800 5.000 olio2\lam_e metal =16/S=8 olio2\lam_g cer =19/S=9 olio2\lam_e cer =20/S=16 olio2\lam_b cer =51/S=0 1.000 olio2\lam_b cer =51/S=0 microstrain, (ue) 10.400 11 3.000 700.000 960.000 1220.000 1480.000 1740.000 2000.000 Eta
Observations on mechanical systems that create Pad Crater Defects The dominant mechanical failure mode in Pb-free compliant materials is Pad Crater and there are no non-destructive methods to identify crack initiation or even significant cracking. The strain rate dependency defined in IPC 9704 should be modified for Pb-free. The board thickness dependency has been validated. Significant differences have been identified in laminate materials. Package stiffness is also a significant factor and could be incorporated into a model generated for a specific program. Primary drivers for a detailed model: Board Thickness / Strain rate / Laminate Material / Package Compliance 12
Pad Crater Mitigation SMD Corner Pads Corner pads are 30.6% larger than NSMD but solderable area is reduced by 9% orce distribution on pad is changed due to change in ratio between pad size and solderable area Reinforcing effect of solder mask has not been quantified Results Board tested to mixed fail range shows failure at IMC. Yellow dye cross section of same tested board shows pad crater traveling below first layer of glass bundle. Crack at IMC will almost certainly propagate to failure under thermal cycling loads. Crack into the first weave presents an unknown risk. 13
Pad Crater Mitigation TV Stack Design with High Modulus Polyimide Layer Weibull Analysis 20 Layer Primary Attach ailure mode converts to IMC interface. No significant increase in survivable strain Beta 40.000 32.200 24.400 16.600 Unreliability (ue) 99.000 90.000 50.000 10.000 Contour Comino_20L_PA_no_poly\Data 1 =20/S=24 90% Comino_20L_PA_poly\Data 1 =15/S=13 90% Probability-Weibull Comino_20L_PA_no_poly\Data 1 =20/S=24 Comino_20L_PA_poly\Data 1 =15/S=13 5.000 8.800 26 layer Primary Attach 1.000 100.000 1080.000 2060.000 3040.000 4020.000 5000.000 1.000 Strain (ue) ailure mode converts to IMC interface. 15% increase in survivable strain 30.000 24.800 99.000 90.000 50.000 Contour Comino 26L_PA_poly\Data 1 =13/S=15 90% Comino_26L_PA_no_poly\Data 1 =38/S=30 90% Probability-Weibull Comino 26L_PA_poly\Data 1 =13/S=15 Comino_26L_PA_no_poly\Data 1 =38/S=30 19.600 However we did find some artifacts in the un-failed / surviving samples Beta 14.400 9.200 10.000 5.000 4.000 800.000 1040.000 1280.000 1520.000 1760.000 2000.000 Eta 1.000 Strain (ue) Brian Gray Celestica Inc 8/15/2011 12:44:13 PM 14
lat Section Results Polyimide Lot Destructive analysis of surviving samples. Separation at the polyimide / epoxy interface is evident in some samples. No connection to atmosphere. There is some possibility that the separation has been increased by the sectioning process. 15
lat Section Results Reference Lot Destructive analysis of surviving samples. We did find an example where separation had occurred between the resin and the reinforcing glass. No connection to atmosphere. This observation must be referenced back to our original assumptions and reservations on crack initiation. 16
Thank you John McMahon P.Eng jmcmahon@celestica.com