The interconnection of circuits required in Large Scale Integration (LSI) using multi-level metalization above monolithic semiconductor arrays is taking basically two approaches. One is predicated on processing with a reasonable yield entire arrays without any semiconductor defects (i.e., 100 percent yield chips) which allows once-generated fixed-wiring patterns to obtain the required interconnect. The second approach aims at much larger semiconductor arrays (i.e., full-slice LSI) for which defect-free processing cannot be expected. Thus, probe tests are made of the semiconductor circuits processed on each LSI slice (or wafer) and record is made of the good and bad circuit positions. Unique interconnection masks are then generated to interconnect good circuits in each wafer's particular yield pattern using certain "discretion" in avoiding the bad circuits. As a result, the 100 percent yield approach emphasizes the need to use standard interconnect masks but is complexity limited by the occurrence of defective circuits in larger arrays, whereas approaches capable of routing around the defective circuits have required a full set of unique signal interconnect masks for each wafer's particular yield pattern.
[1]
Richard L. Petritz.
Current Status of Large Scale Integration Technology
,
1967
.
[2]
J. B. Angell,et al.
Redundancy for LSI Yield Enhancement
,
1967
.
[3]
Don F. Calhoun.
High speed modular multiplier and digital filter for LSI development
,
1968,
AFIPS '68 (Fall, part I).
[4]
L. M. Spandorfer.
Large Scale Integration - an Appraisal
,
1968,
Adv. Comput..
[5]
J. W. Lathrop,et al.
A discretionary wiring system as the interface between design automation and semiconductor array manufacture
,
1967
.
[6]
A.G.F. Dingwall.
High-yield processing for fixed-interconnect large-scale integrated arrays
,
1968
.
[7]
B. T. Murphy,et al.
Cost-size optima of monolithic integrated circuits
,
1964
.