Reliability Monitoring of a Separable Land Grid Array Using Time Domain Reflectometry

Transcription

1 17 th IEEE Workshop on Signal and Power Integrity Paris, France May 12-15, 2013 Reliability Monitoring of a Separable Land Grid Array Using Time Domain Reflectometry Michael H. Azarian (CALCE) College Park, MD USA mazarian@calce.umd.edu

2 Reliability Analysis of Area Array Interconnects Large area array packages: high thermo-mechanical strains Alternatives to ball or column grid arrays: separable land grid array (LGA) connectors employing compressible contacts. Unknown: interconnect failure mechanisms, or degradation of high frequency performance prior to loss of continuity BGA CGA Separable LGA Interposer 2

3 Cross-section of Cinch CIN::APSE Interconnect Solderless contact: 7 to 11 points of contact Randomly wound gold-plated molybdenum wire Hour-glass shaped cavity in polymer substrate 3

4 Approach Develop test vehicle for high frequency signal monitoring of Cinch CIN::APSE separable LGA Set up temperature cycling test: simultaneous monitoring of TDR response and DC continuity (event detector) Perform thermal fatigue testing of test vehicles Compare results of continuity monitoring with changes in TDR response, and document changes over time prior to failure 4

5 Test Board Layout Circuits for TDR Monitoring Ch 1 Ch 2 50 Ohm Term Ch 3 Circuits for Event Detector Integrates TDR and event detector monitoring on the same circuit board 5

6 Layout of TDR Section Impedance: 50 Ohms ± 5 Ohms Input = Pin 1; Output = Pin 16 All other pins are grounded Net lengths between 6.3 and 7.5 cm (2.5 and 3.0 inches). 1 Channel Channel 2 Allows localization of degradation to interconnects with varying distance from the neutral point 5 6

7 Assembly of Test Vehicle Event Detector Circuits Dummy Device 102 mm TDR Circuits Interposer: 7296 I/Os; 320 I/Os populated, in twenty 4x4 arrays 7

8 Close-ups of Interconnect Clusters Continuity TDR Board Dummy Device 1 mm pitch 8

9 Experimental Setup LGA test vehicle (1 of 2) Vector Network Analyzer for TDR Monitoring (45 MHz to 10 GHz) 50 Ohm RF cable Ribbon cable 6-way RF switch (<24GHz) Temperature chamber (-55 to 125 C, 10 C/min ramp) Event detector 40 channels/board 9

10 Continuity Monitoring Analysis Tech 256STD event detector Detection of high resistance events of >200 nsec Resistance threshold: >300 Ohms Each channel monitored 4 interconnects in series No failures detected 10

11 Example of TDR Response Channel 2 after about 620 Cycles, Absolute Calibration Data averaged over all temperatures 11

12 Initial Response at Room Temperature Prior to Temperature Cycling 12

13 Initial Responses at Room Temperature Prior to Temperature Cycling 13

14 TDR Response at Room Temperature after 424 Cycles 14

15 TDR Response at Room Temperature after 424 Cycles 15

16 Correspondence Between TDR Measurements From Either End of the Same Circuit Ch 1 Ch 2 After 424 Cycles, Room temperature 16

17 Effect of Temperature on TDR Response From 484 to 486 Cycles 125 C -55 C Temperature ( C) 17

18 TDR Reflection Coefficient at Room Temperature Channel 2, Extracted from Cycling Data 18

19 Response of Channel 1 Interconnects over 1400 Temperature Cycles Outermost Cluster (#1) Data acquired using absolute calibration 19

20 Observations Event detector: no open circuits over the 1400 cycle test. TDR monitoring: Outermost interconnects on each channel exhibit increased impedance. The connection at the SMA edge connectors also degraded. Temperature has a pronounced effect on the impedance data. It is necessary to evaluate TDR response over a narrow range of temperatures. Multi-channel, automated TDR monitoring over temperature was successfully implemented for reliability monitoring of a complex package with multiple interconnects arranged in a series circuit. 20

21 Conclusions TDR monitoring during thermal cycling of a large separable LGA (with 7296 I/Os) has shown that degradation of high frequency impedance can be significant even in the absence of continuity failures. For applications requiring transmission of high frequency signals, this demonstrates that monitoring only DC resistance or continuity may lead to erroneous conclusions regarding reliability. Contacts located furthest from the center of the package experienced the largest degradation. The variation of TDR response with temperature made it necessary to limit analysis to the temperature of calibration. Future tests could switch between calibrations performed at multiple temperatures. 21

22 Acknowledgements MIT Lincoln Laboratories: Fabrication Engineering James Kelly, Associate Group Leader Roger Maurais, Principal Engineer Student Interns: Juergen Nahm Justus Zeemann Visiting Student: Paolo Manfredi, Ph. D. CALCE Electronics Products and Systems Consortium 22

23 Thank you. 23