Browse Prior Art Database

Latch Design for Improved AC Self Test

IP.com Disclosure Number: IPCOM000122391D
Original Publication Date: 1991-Dec-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 2 page(s) / 78K

Publishing Venue

IBM

Related People

Douskey, SM: AUTHOR

Abstract

An inherent limitation to some Level Sensitive Scan Design (LSSD) functional speed self-test schemes is the inability to guarantee transitions to trigger the test. With pure random patterns these transitions are only accomplished 50 percent of the time. While with weighted random patterns they may be made more likely, there are still many situations that may not be set up with particular hardware designs.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Latch Design for Improved AC Self Test

      An inherent limitation to some Level Sensitive Scan
Design (LSSD) functional speed self-test schemes is the inability to
guarantee transitions to trigger the test.  With pure random patterns
these transitions are only accomplished 50 percent of the time.
While with weighted random patterns they may be made more likely,
there are still many situations that may not be set up with
particular hardware designs.

      This article presents various hardware designs that guarantee
differing values in the L1 and L2 of every latch in the design, thus
allowing the initial trigger transitions at every latch, independent
of consecutive scan latch values.

      A functional speed test utilizes a pair of system clocks in an
LSSD design.  These clocks are run for single or multiple cycles for
AC test coverage.  Latches are initialized through scanning.  The
scan ends on an L1, allowing the L1 to be different than the L2.  The
system clocks are then cycled and the L1s scanned out.  The changing
L2 propagating to the following L1 is the AC test.

      In order for this single cycle test to be AC, the L1 must
differ from the succeeding L2.  Using pseudo-random patterns, this
will only happen half the time.  Thus, with AC coverage lagging DC
coverage, achieving a certain AC coverage may require a
disproportionately large increase in pattern count.

      In a test environment, the L1 and L2 data are under more
control.  Limiting the values to always be opposite allows only two
patterns to be tested:  A's and 5's.  It is the nature of the scan
chain that limits the test set up.

      Allowing the L1 and L2 to differ, even when pseudo-random
patterns are applied, can bridge these limits. Previ...