Professional Documents
Culture Documents
CHAPTER 2:
Input Stimuli: The input stimuli are also based on the specification
Typically, these stimuli correspond to those input and output
specifications that are either critical or considered risky by the synthesis
procedures. A frequently used strategy is to exercise all functions with
only critical data patterns.
Advantages:
1. Its strength lies in the details of the circuit behavior that can be simulated.
For example, logic, timing, and analog behaviors can be simulated.
2.Another advantage is in the use of hierarchy. For example, a design
can be first simulated at a higher behavior level.
Disadvantages:
1. The weakness of this method is its dependence on the designer’s heuristics
used in generating the input stimuli. To contain the complexity, these stimuli are
non exhaustive and, therefore, a guarantee of conformance to specification is
impossible.
Such a guarantee is possible with a formal verification method , which
mathematically proves the correctness of the design. A restricted form of formal
verification,known as model checking verifies finite state concurrent systems by an
exhaustive search of the state-space.
In an electronic circuit, electrical parameters such as voltage and current are used
to encode the information being processed. Thus, in a digital system, a high
6.)Explain Roth’s test detect algorithm
A’)
Faultsimulator needs in addition to the circuit
model, stimuli and expected responses (that are
needed for true-value simulation):
Fault model
Fault list
C(f1)...C(fn) are copies of the defect-free circuit C( ) with fault fx
permanently inserted.
Here, each time the fault is detected, the
simulator records the vector number (and
possibly the output(s) in error).
o Although useful for fault diagnosis,
this is compute expensive.
o Fault dropping causes simulation of
C(fn) to stop after vector 35.
Serial Fault Simulation
If fault dropping is not employed, the effort of
simulating n faults is equivalent to either:
Simulating a circuit n times larger or
Repeating the original true-value simulation n times.
Serial Fault Simulation
o True-value simulation is performed
across all vectors and outputs saved.
o Faulty circuits are simulated one-
by-one by modifying circuit and
running true-value simulator.
o Simulation of faulty circuit stops as
soon as fault is detected.
Adv:
o Any type of fault can be simulated,
e.g., stuck-at, stuck-open, bridges,
delay and analog faults.
For n faults, CPU time can be almost n times
that of a true-value simulator.
o Fault dropping significantly
improves on this.
8.) Explain about parallel fault simulation
Parallel Fault Simulation
Most effective when:
Circuit consists of only logic gates.
Stuck-at faults are modeled.
Signals assume only binary, 0 or 1, values.
All gates have the same delay (zero or unit).
Under these conditions, circuits C(fn) are almost identical.
Here, the bit-parallelism of logical
operations in a computer can be useful.
For a 32-bit word, 1 fault-free and 31 faulty
circuits can be simultaneously simulated.
This yields a speed up of w - 1, with w equal
to the word size.
o If fault dropping is employed,
simulation stops when all w - 1 faults
are detected.
o Therefore, serial fault simulation
has more to gain by fault dropping.
Parallel Fault Simulation
Parallel fault simulation of two faults, c SA0 and f SA1:
Parallel fault simulation cannot accurately model rise and fall delays.
o The signal values in all circuits are
processed simultaneously.
o Zero-delay or unit-delay are used.
Compiled-code or event-driven versions are
possible.
o Multi-valued logic is possible, e.g.,
(0, 1, X and Z), by encoding state in
more than 1 bit.
A true-value logic simulator can be used as a
parallel fault simulator by inserting gates to
model faults -- see text.
Parallel Fault Simulation
With ab = 10, only L b is sensitizable to f (faults on a are masked).
o The faults given by
o Had b = 1, L a would have been sensitized to f , e.g.,
11.) Explain about differential fault simulation
A.) Cheng
and Yu made 2 improvements on
TEST-DETECT.
Eliminated the use of D-calculus.
Eliminated the explicit restoration to true-value before processing the next
fault.
The algorithm, which starts with a vector set
and a fault list:
Simulate a vector in true-value mode and store the PO values.
Activate a fault by creating a transition to the faulty value, e.g., if true-value is
0 and it is a SA1, generate a 0 -> 1 transition.
Simulate the circuit and check for a difference at POs -- drop the fault if
detected.
For next fault, restore to true value by placing a restoring transition at previous
site. Place a second transition at new fault site and simulate.
Repeat with the next vector once all faults have been analyzed.
Figure 5.23: Differential fault simulation example: first vector (1,1).
12.) Explain why the reverse-order fault simulation is not a practical test compaction
technique for sequential circuits.
A.)
Fault simulation: Consider a four-bit ripple-carry adder circuit of
the type shown in Figure 5.2. This circuit contains four full-adder blocks of Figure
5.2(a.) It has 36 logic gates, 9 primary inputs, and 5 primary outputs. A fault
simulator [160] makes a list of 186 single stuck-at faults. The equivalence fault
collapsing (see Chapter 4) option in the simulator reduces it to 114 faults. The
result of fault simulation with vectors of Table 5.1 is shown in Table 5.2. We make
several observations:
• The last vector does not detect any new fault. We can use the first six vectors for
testing the circuit without any reduction in the fault coverage. Even though the last
two vectors were necessary for the design verification heuristic of the last section,
they are found to be redundant for testing. For large circuits, it is not unusual to
have several million vectors in the verification vector set. A fault simulator can
significantly reduce the size of the vector set, thus reducing the time required for
the manufacturing test.
• The collapsed fault set is about 40% smaller than the set of all faults. A 40-50%
reduction in the fault list is quite typical.
• Although the coverages of collapsed and uncollapsed fault lists differ, they are
quite close. The CPU time for simulation of all faults on a SUN Sparc 2
computer was 33 ms, and reduced to 16 ms for the collapsed fault set. For
large circuits, the saving in the CPU time may be significant.
Note that vector 6 is essential since it detects some faults that could not be detected by all 5
previous vectors. It is, however, possible that vector 6 could detect faults covered by some
previous vectors, making those vectors unnecessary. This could not be determined earlier
because of “fault dropping” by the simulator. That is, a fault detected by a vector is immediately
dropped from consideration and is not simulated for the following vectors. So, we will simulate
vector 6 first. Similarly, we can argue that vector 5 could also detect faults covered by some
earlier vectors. A simple way to find that out is to simulate the vectors in the reverse order. The
result is shown in Table 5.3. We can now use four vectors above the line in this table for a 100%
coverage. Unfortunately, the reverse-order simulation cannot be used for sequential circuits.
13.) Explain what action an event-driven true-value simulator will take when it evaluates
a zero-delay gate?
A.)