Browse Prior Art Database

High Reliability Voter Configuration

IP.com Disclosure Number: IPCOM000101714D
Original Publication Date: 1990-Aug-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 2 page(s) / 50K

Publishing Venue

IBM

Related People

Aichelmann, FJ, Jr: AUTHOR

Abstract

A method is proposed for providing an improved reliability version of triple-modular-redundance (TMR) systems. This development makes it possible to cover the voter by extending error-checking-correction (ECC) through the voter network.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 84% of the total text.

High Reliability Voter Configuration

       A method is proposed for providing an improved
reliability version of triple-modular-redundance (TMR) systems.  This
development makes it possible to cover the voter by extending
error-checking-correction (ECC) through the voter network.

      In TMR applications a reliability improvement is obtained over
conventional simplex data path by voting two out of three elements in
parallel.  Considerable self-checking is included within these voter
networks to insure correct operations.  However, the voter network is
subject to failures even with self-checking circuits because they are
unable to correct failures.  Once the voter fails, the whole
configuration is unreliable.  This technique enables the voter
network to be correctable in TMR applications that have ECC within
the elements (or paths). These are voted by passing encoded data
through the voter. The results can then be corrected by a
conventional ECC facility.

      Fig. 1 describes a typical TMR application where three elements
are voted to determine a majority (i.e., 2 out of 3).  Fig. 2 is an
example of a system that has ECC between the elements and the voter
and passes the ECC checking and correction by voting encoded data
through the voter.  As a result, this configuration can improve the
reliability by being able to correct failures in the voter.  Memory
systems already have encoded data within the memories and, therefore,
would not incur overhead p...