Test-Architecture Optimization and Test Scheduling for TSV-Based 3-D Stacked ICs

Authors:
Brandon Noia;Krishnendu Chakrabarty;Sandeep Kumar Goel;Erik Jan Marinissen;Jouke Verbree
Affiliations:
Department of Electrical and Computer Engineering, Duke University, Durham, NC, USA;Department of Electrical and Computer Engineering, Duke University, Durham, NC, USA;TSMC Technology, Inc., San Jose, CA, USA;3-D Integration Program, IMEC, Leuven, Belgium;Department of Computer Engineering, Delft University of Technology, Delft, The Netherlands
Venue:
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Year:
2011

Citing 0
Cited 2

Post-bond stack testing for 3d stacked IC

VDAT'12 Proceedings of the 16th international conference on Progress in VLSI Design and Test
A thermal-driven test application scheme for pre-bond and post-bond scan testing of three-dimensional ICs

ACM Journal on Emerging Technologies in Computing Systems (JETC)

Quantified Score

Hi-index	0.03

Visualization

Abstract

Through-silicon via (TSV)-based 3-D stacked ICs (SICs) are becoming increasingly important in the semiconductor industry. In this paper, we address test architecture optimization for 3-D stacked ICs implemented using TSVs. We consider two cases, namely 3-D SICs with die-level test architectures that are either fixed or still need to be designed. We next present mathematical programming techniques to derive optimal solutions for the architecture optimization problem for both cases. Experimental results for three handcrafted 3-D SICs comprising of various systems-on-a-chip (SoCs) from the ITC'02 SoC test benchmarks show that compared to the baseline method of sequentially testing all dies, the proposed solutions can achieve significant reduction in test length. This is achieved through optimal test schedules enabled by the test architecture. We also show that increasing the number of test pins typically provides a greater reduction in test length compared to an increase in the number of test TSVs. Furthermore, we show that shorter test lengths are generally achieved with the larger, more complex dies lower in the stack. This is because test data must pass through every die lower in a stack in order to reach its target die, and with the larger dies lower in the stack, more test bandwidth may be provided to these dies using fewer routing resources.