TDEB Top Bar

Trending
Semiconductor Industry News — The Digital Electronics Blog
Follow Us

TDEB Masthead

The Semiconductor Hub

Semiconductors.  Chips and Systems.  Innovation.

Indexed by leading Search Engines, AI and referenced by Engineers worldwide 500+ Articles 10M+ Pageviews 30+ Reports 50+ Citations

Most Significant Papers on IC Test (ITC)

Aarthi Chinappa
by
0
Share Facebook X LinkedIn WhatsApp Email

Many important papers have been presented at the annual International Test Conference (ITC) since it began in 1970. Each year ITC picks the most significant paper it published ten years earlier, and the full list is on the ITC Significant Papers page. This post explains what ITC is, lists the award winners up to 2016 with a line on what each is about, and adds a deep dive on the test concepts behind them.

What ITC is

ITC is the main conference for electronic test of devices, boards and systems, covering design for test, test generation, diagnosis, failure analysis and feedback into process and design. According to ITC's history page, it started in 1970 as a symposium on testing the then-new semiconductor memory chips, held at the Rickshaw Inn in Cherry Hill, New Jersey, with 147 people. It is now a week-long event with sister conferences in India and Asia.

For anyone working on DFT, ATE, yield or silicon debug, the ITC proceedings are a primary record of how test methods developed.

The Most Significant Paper award

Each year ITC chooses the most significant paper published at the conference ten years before, judged on impact and significance, relevance, and historical interest.

Award winners, published 1998 to 2006

From the ITC Significant Papers page; the one-line notes are ours.

Published / awardedPaperWhat it is about
1998 / 2008A Structured and Scalable Mechanism for Test Access to Embedded Reusable Cores (E.J. Marinissen and others)How to reach and test IP cores buried inside an SoC
1999 / 2009Defect-based Delay Testing of Resistive Vias-Contacts: A Critical Evaluation (K. Baker and others)Whether delay tests catch resistive via and contact defects
2000 / 2010Wrapper Design for Embedded Core Test (E.J. Marinissen, M. Lousberg, S. Goel)Designing the test wrapper around a core to cut test time
2001 / 2011OPMISR: The Foundation for Compressed ATPG Vectors (C. Barnhart and others)Compressing scan test data and responses
2002 / 2012Embedded Deterministic Test for Low-Cost Manufacturing Test (J. Rajski and others, Mentor Graphics, Poznan University of Technology, Cisco)On-chip decompression of deterministic scan patterns
2003 / 2013A Case Study of IR-Drop in Structured At-Speed Testing (J. Saxena and others, Texas Instruments, Siemens)Supply droop during at-speed scan tests causing false failures
2004 / 2014Scan-based Side-Channel Attack on Dedicated Hardware Implementations of Data Encryption Standard (B. Yang, K. Wu, R. Karri)How scan chains can leak secret keys
2005 / 2015Invisible Delay Quality Lights Up What Could Not Be Seen (Y. Sato and others)Measuring how well delay tests cover small delay defects
2006 / 2016Preferred Fill: A Scalable Method to Reduce Capture Power for Scan Based Design (X. Lin, I. Pomeranz, J. Rajski and others)Choosing don't-care bits to lower power during scan capture

The ITC page currently lists awards up to 2016.

Other useful ITC pages

Deep dive: the numbers and details

This part holds the conference figures and worked examples of the test concepts behind the award winners. Skip it if the overview is all you need.

How selective ITC is

From ITC's acceptance rates page:

YearRegular papers acceptedRate
202330 of 16218.5 %
202436 of 15523.2 %
202540 of 16424.4 %

The ITC 2027 call for papers asks for complete papers of up to 10 pages, notes that submissions under 4 pages are rarely accepted, and sets title and abstract deadline of March 5, 2027 for the conference on October 10 to 15, 2027 in San Antonio, Texas.

Why scan compression matters: a worked example

Two of the winners (OPMISR and Embedded Deterministic Test) are about test compression. The arithmetic shows why. Take a design with 100,000 scan flip-flops and a tester with 10 scan channels (illustrative numbers):

Without compression: 10 chains of 10,000 cells
  shift cycles per pattern = 10,000

With on-chip decompression: 10 tester channels feed 500 internal chains
  chain length = 100,000 / 500 = 200 cells
  shift cycles per pattern = 200

Reduction in test time and data per pattern = 10,000 / 200 = 50x

At a 50 MHz shift clock, 5,000 patterns take 5,000 x 10,000 / 50e6 = 1.0 s without compression and 0.02 s with it. Tester time is paid per second per device, so this saving goes straight into the cost of every chip.

Why coverage targets are so high

A widely used model relates the defect level (DL, the fraction of shipped parts that are bad) to process yield Y and fault coverage T:

DL = 1 - Y^(1 - T)

Y = 0.90, T = 0.99:   DL = 1 - 0.90^0.01  = 0.00105 = about 1,050 DPPM
Y = 0.90, T = 0.999:  DL = 1 - 0.90^0.001 = 0.000105 = about 105 DPPM

Going from 99 to 99.9 percent coverage cuts escapes by about ten times.

Scan power and IR drop

Two winners (the 2003 IR-drop case study and Preferred Fill) deal with power during test. In functional mode only a fraction of flip-flops toggle each cycle, but scan shift and capture can toggle far more at once. More switching means more current and more supply droop, which slows gates and can make a good chip fail an at-speed test. Filling the don't-care bits of test patterns so that fewer flip-flops change state is a low-cost way to reduce that switching.

Share Facebook X LinkedIn WhatsApp Email

Post a Comment

0 Comments

Your comments will be moderated before it appears here.

Post a Comment (0)

#buttons=(Ok, Go it!) #days=(20)

Our website uses cookies to enhance your experience. Learn more
Ok, Go it!