Diagnostic Wrap/Assign Spare Functions for a Crosspoint Switch
Original Publication Date: 1991-Nov-01
Included in the Prior Art Database: 2005-Apr-04
Publishing Venue
IBM
Related People
Bjornseth, PJ: AUTHOR [+2]
Abstract
A matrix switch design based on the double-sided matrix configuration (see Fig. 1) allows connectivity between each input and each output through a single crosspoint. Due to the lack of any redundancy inherent in the design, as compared to a single-sided configuration, a method of preventing the loss of connectivity between any input-output combination is desirable.
Diagnostic Wrap/Assign Spare Functions for a Crosspoint Switch
A matrix
switch design based on the double-sided matrix
configuration (see Fig. 1) allows connectivity between each input and
each output through a single crosspoint.
Due to the lack of any
redundancy inherent in the design, as compared to a single-sided
configuration, a method of preventing the loss of connectivity
between any input-output combination is desirable.
Based on the
implementation of a highly reliable CMOS
technology, the case of single internal failures was examined. It
was found that the INMUX/REG2 stages of the design (see Fig. 2)
required approximately 34% of the internal cells used to construct
the entire matrix design. After thorough analysis of the overall
design, it was determined that the INMUX/REG2 stages would be the
ideal area to add redundancy to the matrix design. Ease of
implementation and minimization of additional logic (cells) required
to achieve a high degree of fault-tolerance were the reasons for this
decision.
Two
operations are used for implementing the spare logic in
place of a destination's normal INMUX/REG2 logic, and for testing the
switch's data paths. A command to
instruct the switch to reassign
the spare data path for a destination's normal data path may be added
to the switch controller's other functions.
After receiving the
"assign spare column" command, the switch will substitute the spare
INMUX/REG2 stages for the INMUX/REG2 stages normally driving the
destination whose address is given with the "assign spare" codepoint.
This ef fectively bypasses a faulty INMUX/REG2 path through the
redundant logic. The spare column will
be used from that point on
during any operation involving the destination involved in the
"assign spare" command.
In order to
test for an internal fault, a "diagnostic wrap"
function is used which drives a test input's data through a data path
established by the switch controller. This "wrapped" data is gated
into REG1, passes through the crosspoint established by the switch
controller,...