Simulation
Last Edit July 22, 2001
Wafer Sort/Packages Part Sort Functional Simulation
The vectors used to perform wafer sort and packaged part testing are
generated from simulation output vectors. These simulation output
vectors consist of all the input and the expected output signals generated
by the input file. These vectors provide a time independent, sequential
state description of the circuit after any input change has propagated
through the logic and all the outputs have settled to their stable state.
A reasonable sample step for this simulation is 100ns, with the sample
taken one simulator time step before the next input vector.
The simulation provided to the array vendor should contain sufficient
vectors to provide verification of the logic and should supply all simulation
vectors required for the recommended 90% or better fault coverage of the
final circuit. The level of fault coverage will vary from vendor to vendor
or may be dictated by the designer's company. Indications are that reliability
requirements are beginning to force a high (98-99%) fault coverage requirement.
Fault-Grading
Fault grading is a measure of the fault coverage - a "grade" on
the quality of the fault detection provided by the submitted simulation
vectors. A fault grading score of 100% means that if a SA1 or SA0 fault
exists at any single observable node within the circuit it will be detected
during the tester functional testing phase.
Single fault detection, the detection of a stuck-at fault (SA1 = stuck-at-1;
SA0 = stuck-at-0) at any single node in the circuit, requires that the
node be "covered" by at least one simulation vector. A node is covered
by the vector set when the state of at least one circuit primary output
for at least one vector is different when the failure is present than
when the failure is absent. A failure at a circuit node that is not
covered by the functional simulation vector set will not be detected.
A stuck-at fault is generally thought of as a physical open or short
circuit, i.e., a "hard" failure. Intermittent failure is not necessarily
detectable by the stuck-at model.
Redundancy in the circuit produces fault-masking and will reduce the
obtainable fault-coverage since it reduces the observable nodes. The addition
of test points when the redundancy is deliberate, and the minimization
of the circuit when it is not, are recommended approaches to improve testability.
There are no requirements for fault location, i.e., the identification
of the exact point of failure. Multiple-fault detection, a less-probable
occurrence, is also not required although most single-fault minimal test
sets and minimal test sequences will provide 100% fault coverage of all
observable faults and will also detect the presence of many multiple faults,
depending on the circuit implementation.
The Minimal Test Sequence as applied to combinatorial and sequential
circuit elements is discussed in Chapter
9.
Table 8-1 Questions To Be Asked
Questions to Ask About Simulation Submission Requirements |
- What simulations are required fordesign submission and on what
media?
- wafer-sort
- timing verification at speed
- AC tests
- parametric tests
- other
- Electronic FTP, floppy disk or tape?
- What simulations may be submitted?
- What simulators are supported?
- What support for hazard and race detection?
- What timing verifiers are supported?
- What level of fault-grading is required?
- by the array vendor
- by the designing company
- What rules must be followed for each type of simulation
- Determine if ATG outputs are acceptable or if they need a post-processing
- What is submitted?
- simulator control or command files
- input files
- output files
- vector checking reports
- netlist
- annotation delay files
- other control or data files?
- Is annotation software available?
- What testers are available?
- What expansion can be expected from the test program?
- What vector-checking software is available?
|
Testability Analysis
There were several available software packages such as the Dazix DTA
(Dazix Testability Analyzer) and Tegas COPTR that analyzed the testability
of a design. They attempted to measure the controllability and the observability
of the nodes within the design. There are various products on individual
workstations that have been designed to perform this task. Note that array
vendors on the whole do not enforce their use and may not support them
if different models are required.
Controllability is the measure of how difficult it is to set a node
to a given value. A node is controllable if it takes one or not more
than a selected number of vectors to set it to a given value, i.e., propagate
a primary input signal to that node. The ideal case is that it requires
one vector to set a node.
Observability is the measure of how difficult it is to see the value
to which a node is set. A node is observable if it takes one or not
more than a selected number of vectors to propagate the value of the node
to an observable output. The ideal case is that it requires one vector
to observe a node.
Design optimization using ad hoc or formal procedures (DFT) to improve
testability scores were discussed earlier. If such software is available,
it is recommended that the design evaluate the circuit for testability
before finalizing the design and proceeding with the simulations.
|