8. 114.2857143
85.71428571
102.1978022 95.81780538
minset RV1 RV2 RV3
c17
311.7346939
284.3537415 284.0732169
276.9742498
minset RV1 RV2 RV3
c432
1603.755102
1555.265306
1547.737374 1549.625651
minset RV1 RV2 RV3
c499
1033.936652
950.2262443
933.9992532
950.1333926
minset RV1 RV2 RV3
c880
1589.859944
1547.787115
1552.794988 1550.491513
minset RV1 RV2 RV3
c1355
1251.645065
1192.422732 1187.582881
1201.120637
minset RV1 RV2 RV3
c1908
Part(c)Data and Conclusions Min set –Minimum set
generated by Atlanta.(size
same as mentioned in part a)
RV1-Random test vector set
with size same as the minset
RV2-Random test vector set
with 2x size of minset
RV3-Random test vector set
with 10x size of minset.
Y-axis-Average hamming
distance percentage.
Average hamming distance
=Sum of gamming distance
of all pairs/total no.of pairs.
Average hamming distance
percentage=Average
hamming distance*100
9. 6432.650494
6171.015274 6182.044174 6193.045059
minset RV1 RV2 RV3
c2670
946.4447543
676.4726386 682.8274275 678.2226881
minset RV1 RV2 RV3
c3540
5189.772341
5018.722392
4967.276878 4977.886454
1 2 3 4
c5315
1562.388592
1523.707665 1521.817384 1519.835155
minset RV1 RV2 RV3
c6288
4970.456503
4865.284791
4879.928197
4900.259947
minset RV1 RV2 RV3
c7552 When compared to the Atalanta generated test vectors the random
vectors have a lower hamming diversity in case of all benchmarks. This
means the random vectors considered produced are such that the
outputs have smaller difference between each other than the outputs
produced by Atalanta.As Atalanta covers all the possible faults it might
cover the entire range of outputs in order to see the difference and
hence has more diversity.
10. From the graphs it can be seen that enlarging the test pattern did not give a significant difference in the hamming distance
diversity in most of the cases(but not all).What I understood from this is the random vectors I have chosen might be in such a
way that they are leading a change in the output of only a specific part of the circuit .
So in order to increase the HD diversity the test vectors used to increase the random set should be in such a way that they
can produce noticeable change across the output. For example for C17 benchmark the Random test set with double size than
the minset shows greater diversity than the Random test with size similar to that of minset .This means the increased vectors
are capable of generating wide change in output values leading to increased diversity. It is similar in case of C7552 benchmark
also.
So from this we can conclude that the HD diversity depends on the variation that the input test set can make at the output .
If Inputs in the random set are increased as shown in Figure a i.e. addition of these inputs can change large number of
outputs then the diversity increases( with exceptions -what if they are covering the same part of circuit as inputs that are
already existing?? In such a case the diversity does not change much.)
If inputs are as shown Figure b i.e. addition of inputs cover a smaller part of output then the diversity decreases.
One more observation is the diversity depends not only the type but also depends on size because as the number of
combinations increase the denominator for average hamming distance( which we are considering as a measure for diversity )
increases and hence the diversity decreases .So the set size should be increased as less as possible with as much output
coverage as possible to increase the diversity.
Figure a Figure b