Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
9
33
~
940
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
2
.i
3
.pp
9
33
-
9
40
933
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
A SoC
-
I
P Core
Test D
ata Compr
ession Sc
heme
B
ased
on Err
or
Corre
ctin
g Ham
min
g Cod
es
Sa
n
joy Mitr
a
1
,
De
bapras
ad
Da
s
2
1
Depa
rtment of
Com
pute
r
Scie
n
ce
and Engi
ne
ering,
Tr
ipura Insti
t
ute
of
T
ec
hnolo
g
y
,
Agart
ala, Ind
ia
2
Depa
rtment of
El
e
ct
roni
cs
and
Com
m
unic
at
ion Engi
ne
eri
ng,
TS
SO
T,
As
sam
Univer
sit
y
,
Sil
cha
r
,
India
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
26
,
201
8
Re
vised
Ma
y
2
1
, 2
01
8
Accepte
d
Aug
19
, 201
8
As
sy
st
em
-
on
-
c
hip
(SoC)
in
te
g
rat
ion
is
growi
ng
ver
y
r
api
d
l
y,
in
creased
ci
rcu
it
densities
in
SoC
have
le
a
d
a
rad
ical
in
crease
in
te
st
da
ta
volume
and
red
uction
of
th
is
la
rge
t
est
data
v
olume
is
one
of
t
he
bigge
st
ch
al
l
e
nges
in
th
e
te
sting
industr
y
.
Th
is
pap
er
pre
sents
an
eff
i
cient
te
st
inde
p
ende
nt
compress
ion
sche
m
e
primarily
b
ase
d
on
the
err
or
cor
re
cting
Ham
m
ing
code
s.
The
sche
m
e
oper
ates
on
the
pre
-
computed
te
s
t
dat
a
withou
t
t
he
nee
d
of
struct
ura
l
informati
on
of
th
e
ci
r
c
uit
under
t
est
an
d
thus
it
is
appl
i
ca
bl
e
for
IP
cor
es
in
SoC
.
Te
st
ve
ct
ors
ar
e
equall
y
sli
ce
d
int
o
th
e
siz
e
of
‘
n’
bit
s
.
Indivi
dual
slices
are
treate
d
as
a
Ham
m
ing
code
word
consisti
ng
of
‘
p’
par
i
t
y
bit
s
and‘
d’
data
bit
s
(
n
=
d
+
p
)
and
val
id
ity
of
e
ac
h
code
word
is
ver
ifi
ed
.
If
a
val
id
sli
ce
is
enc
ount
ere
d
‘
d’
dat
a
bit
s
pre
f
ix
ed
b
y
‘
1’
ar
e
writt
en
to
the
compress
ed
fil
e
,
while
for
a
non
-
val
id
sli
ce
al
l
‘
n’
bit
s
pre
c
ede
d
b
y
‘
0’
ar
e
writt
en
to the
co
m
pre
ss
ed
fil
e.
Fi
nal
l
y
,
we
apply
Huffm
an
codi
ng
and
RLE
i
n
orde
r
to im
prove
the
compress
ion
rat
io
furth
er
Th
e
eff
ic
i
ency
of
th
e
proposed
h
y
br
id
sche
m
e
i
s
ver
ified
with
t
he
exp
eri
m
ental
outc
om
es
and
c
om
par
isons
to
ex
isti
ng
compress
ion
m
et
hods
suita
ble for
te
st
i
ng
of
IP
cor
es.
Ke
yw
or
d
s
:
Com
pr
ession
Ham
m
ing
code
Huff
m
an
co
ding
Non
-
valid
sli
ce
RLE
Test
d
at
a
Vali
d
sli
ce
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Sanjo
y M
it
ra,
Depa
rtment
of
Com
pu
te
r
Scie
nce a
nd E
ng
i
ne
erin
g,
Trip
ur
a
Insti
tute o
f
Tec
hnolog
y,
Nar
si
ng
a
r
h Agart
al
a, Tr
i
pura
(
W)
,
India.
Em
a
il
:
m
ai
l.s
m
it
ra@g
m
ai
l.com
1.
INTROD
U
CTION
The
pr
im
e
obj
e
ct
ive
of
te
st
da
ta
com
pr
essio
n
is
to
le
sse
n
th
e
volum
e
of
bi
nar
y
bits
in
or
i
gin
al
ATPG
gen
e
rated
te
st
cub
e
.
is
sto
re
d
in
the
A
n
a
ut
om
atic
te
st
equ
ipm
ent’s
(
AT
E)
inte
rn
al
m
em
or
y
is
us
ed
t
o
sto
re
the
com
pr
esse
d
te
st
vectors
and
a
n
on
-
chi
p
dec
om
pr
ession
hard
war
e
i
s
app
li
ed
t
o
de
com
pr
ess
this
ATE
store
d
data
wh
ic
h
is
s
ub
seq
uen
tl
y
ap
pl
ie
d
to
ci
rc
ui
t
unde
r
te
st
(CU
T)
[1
]
,
s
how
n
in
Fig
ur
e
1.
The
dec
om
pr
ession
ha
r
dw
a
re
is
sp
eci
fical
ly
desi
gn
e
d
for
any
disti
n
ct
c
om
pr
ession
a
pp
ro
ac
h
a
nd
is
s
uitable
for
al
l t
he o
rigi
nal test
set w
hi
ch usin
g
t
his c
om
pr
essio
n
a
ppr
oac
h.
1
.
1.
B
ackgro
und
The
te
st
data
com
pr
essio
n
m
et
hods
m
ay
ordina
rily
be
gro
up
e
d
in
th
ree
ty
pes:
cod
e
-
ba
s
ed
schem
es,
li
near
-
dec
om
pr
ession
-
based
s
chem
es
and
br
oad
ca
st
-
sca
n
-
base
d
sc
hem
es
[
2].
Co
de
-
bas
ed
sc
hem
es
m
os
tl
y
ta
rg
et
the
giv
e
n
te
st
set
s,
in
wh
ic
h
ori
gi
nal
te
st
data
are
bro
ken
i
nto
diff
e
ren
t
sym
bo
ls
an
d
each
sy
m
bo
l
is
su
bst
it
uted
by
a
cod
e
wor
d
to
fo
rm
the
com
pr
ess
ed
te
st
da
ta
.
Her
e,
pr
i
or
knowle
dge
of
the
inter
nal
struc
ture
inf
or
m
at
ion
of
the
ci
rcu
it
unde
r
te
st
(CUT
)
i
s
not
nee
ded
;
be
sides,
th
e
fa
ult
si
m
ulati
on
an
d
te
st
ge
ner
at
io
n
ar
e
no
t
re
qu
ir
ed
.
T
hu
s
,
t
hese
sc
he
m
es
are
es
pecial
ly
handy
f
or
te
st
data
com
pressi
on
with
S
oC
I
P
c
or
e
ci
r
cuits.
These
co
ding
m
et
ho
ds
ca
n
be
cat
egorized
into
two
di
fferent
cl
asses
base
d
on
the
diff
e
re
nces
of
sy
m
bo
l
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
9
33
–
9
40
934
div
isi
on.
In
‘fi
xed’
cat
eg
or
y
,
fixe
d
num
ber
s
of
in
pu
t
bits
are
enc
od
e
d
by
the
unde
rly
ing
c
om
pr
essi
on
m
echan
ism
.
An
al
ogously
in
‘v
a
riable’
ca
te
gory;
var
ia
bl
e
nu
m
ber
s
of
in
pu
t
bits
are
enc
oded
by
the
com
pr
essio
n
al
gorithm
.
Hu
f
f
m
an
cod
in
g
is
the
idle
instanc
e
of
su
c
h
“fixe
d”
schem
e.
The
Hu
f
fm
an
enco
di
ng
al
gorithm
enc
od
e
s
fr
e
quentl
y
occu
r
rin
g
s
ym
bo
ls
with
sh
ort
er
c
od
e
words
an
d
on
the
ot
her
hand
,
le
ast
fr
eq
ue
nt
on
e
s
are
assi
gned
relat
ively
lo
ng
e
r
c
od
e
.
Ot
he
r
instance
s
of
su
c
h
‘
fixe
d’
ca
te
gory
are
sel
e
ct
ive
Huff
m
an
co
di
ng
(SHC)
[
3]
and
opti
m
a
l
sel
ect
ive
Huff
m
an
c
od
i
ng
(OS
HC)
[4
]
,
dicti
onary
-
base
d
c
odin
g
[
5]
and
b
loc
k
m
er
ging
co
ding
[
6]
and
so
on.
Si
ng
le
r
un
-
le
ngt
h
an
d
double
r
un
-
le
ngth
e
ncodin
g
m
e
tho
d
fa
ll
in
the
cat
egory
of
‘
va
riable’
sc
hem
e.
Ru
ns
of
‘
0’
s
are
enc
oded
i
n
case
of
sin
gl
e
run
-
l
e
ng
t
h
c
od
i
ng
te
c
hn
i
ques
an
d
exam
ples
of
t
hi
s
inclu
de
Go
l
om
b
cod
e
[
7],
fr
e
qu
e
ncy
-
dire
ct
ed
r
un
-
le
ngth
(FDR)
c
ode
[
8]
an
d
var
ia
ble
input
Huff
m
an
co
de
(VIH
C
)
[
9].
I
n
case
of
doub
le
run
-
le
ngt
h
c
od
e
c
om
pr
essi
on
te
c
hn
i
ques,
bo
t
h
r
uns
of
0
s
an
d
runs
of
1
s
a
re
en
co
de
d.
E
xtend
e
d
F
DR
co
de
(EFD
R
)
[
10
]
al
te
rn
at
ing
r
un
-
le
ng
t
h
co
ding
(
AFDR)
[
11
]
and
m
ixed
do
ub
le
run
-
le
ngth
and
Huff
m
an
co
ding
(RL
-
HC) [
12]
are
the e
xam
ples
of
double
run
le
ng
t
h
e
nc
od
i
ng
1
.
2
.
Pr
ob
le
m
Faste
r
dev
el
opm
ent
of
In
te
grat
ed
Ci
rcu
it
fa
br
ic
at
io
n
pr
oc
ess
f
or
ci
ng
a
n
inevita
ble
in
c
rease
in
t
he
densi
ty
of
ci
rc
uit
com
po
ne
nt
s
in
a
chi
p
a
nd
this
has
rais
ed
te
st
data
volum
e
a
lot,
w
hich
f
ur
the
r
not
only
enlar
ges
the
te
sti
ng
ti
m
e
bu
t
al
so
surpa
sses
the
te
ste
r
m
e
mo
ry
capa
ci
ty
[1
3]
.
Syst
em
In
te
gr
at
ors
face
s
so
m
e
diff
ic
ulti
es
w
hi
le
te
sti
ng
I
nte
ll
ect
ual
prop
e
r
ty
(I
P
)
c
or
es
a
s
the
st
ru
ct
ur
e
of
the
IP
s
is
unkn
own
t
o
th
e
m
an
d
this
com
plexity
of
IP
c
ores
and
their
siz
e
is
the
pri
m
e
c
ause
of
la
rg
e
r
te
st
data
volu
m
es
and
obvi
ou
sly
,
longer
te
st
ap
pl
ic
at
ion
tim
e
(TA
T
)
is
need
e
d
f
or
qu
al
it
y
post
-
pro
du
ct
io
n
te
st
.
A
la
rg
e
vo
l
um
e
of
te
st
data
is
need
e
d
to
be
s
tore
d
in
the
au
tom
a
ti
c
te
st
equ
ipm
ent
(ATE
)
a
nd
tra
ns
m
itt
ed
dee
p
int
o
th
e
chi
p
as
quic
kly
as
po
s
sible.
Lim
i
t
ed
the
siz
e
of
m
e
m
or
y
and
c
on
sta
nt
cha
nn
e
l
capaci
ty
of
ATE
trig
ge
rs
s
ign
ific
a
nt
rise
in
the
te
st
app
li
cat
ion
tim
e
and
the
te
st
po
wer
.
Te
st
data
co
m
pr
e
ssion
te
ch
niqu
es
hav
e
the
po
te
ntial
to
reso
lute
the
pro
blem
o
f
higher
test
d
at
a
vo
lum
e d
ur
in
g S
oC
-
IP
test
in
g.
1
.
3
.
S
olu
tion
In
t
his
pap
e
r,
error
c
orrecti
ng
Ham
m
i
ng
c
od
e
s
are
ap
plied
f
or
te
st
dat
a
com
pr
essio
n.
Alth
ou
gh
Ham
m
ing
co
de
s
[14,
15]
are
m
os
tl
y
app
li
ed
for
er
r
or
correct
io
n,
it
can
al
so
be
app
li
ed
in
te
s
t
data
com
pr
essio
n
al
lowing
a
sm
all
a
m
ou
nt
of
dist
or
ti
on
[
16
–
19
]
.
In
this
pa
pe
r,
w
e
intr
oduce
a
hybr
i
d
com
pr
e
ssion
schem
e
wh
ic
h
is
pr
im
aril
y
base
d
on
t
he
error
co
rr
ect
i
ng
Ham
m
ing
cod
e
.
We
cl
ai
m
on
the
ba
s
is
of
exp
e
rim
ental
o
utcom
es that t
he
sch
em
e eff
ic
ie
ntly
co
m
pr
ess
es the
So
C
-
IP
Core
te
st
data.
Figure
1
.
S
oC
Test
Mo
del
Figure
2
.
ty
pical
sli
ci
ng
of Te
st Data
Co
m
p
r
ess
ed
T
est
Data a
t
AT
E
On
-
ch
ip
Deco
m
p
r
ess
o
r
Scan
Ch
ain
s o
f
vario
u
s IP
Co
res
On
-
Ch
ip
Co
m
p
a
ra
to
r
Sav
ed
co
m
p
ressed
Go
ld
en
r
esp
o
n
ses
Co
m
p
a
rator
A
T
E
S
I
N
K
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A SoC
-
IP
C
or
e
Test Dat
a
Co
mp
re
ssio
n Sc
he
me base
d o
n Error
Correcti
ng Ha
mm
i
ng Cod
e
s
(
Sanj
oy M
it
ra
)
935
2.
THE
HAM
MI
NG CO
DE B
AS
ED
DAT
A CO
MP
RESSI
ON ALG
ORIT
HM
The
er
ror
-
c
orr
ect
ing
co
de
in
cepted
by
Ha
m
m
ing
find
s
m
or
e
extensiv
e
app
li
cat
ion
i
n
ad
va
nce
d
inf
or
m
at
ion
processin
g
a
nd
c
omm
un
ic
at
ion
syst
e
m
s.
It
is
well
equ
i
pped
for
disti
ng
uish
i
ng
t
wo
bits
er
r
or
a
nd
correct
s
a
sin
gle
bit
error.
Fu
rt
her
m
or
e,
burst
er
rors
[
14,
15
]
can
li
ke
wise
be
c
orr
ect
ed
with
the
ai
d
of
Ham
m
ing
co
de
s
.
Let
us
ass
um
e
a
m
essage
with
d
data
bits
and
it
is
t
o
be
co
de
d
usi
ng
Ham
m
ing
cod
e
s.
The
pr
im
e
idea
o
f
the
Ham
m
i
ng
co
des
li
es
i
n
the
util
iz
at
ion
of
a
ddit
ion
al
pa
rity
bits
(p)
kee
ping
in
m
i
nd
the
end
goal
to
rec
ognize
a
si
ngle
bit
a
nd
an
ide
ntific
at
ion
of
two
bits
e
rror
s
.
Her
e
,
the
‘
n’
bi
ts
of
c
oded
m
essag
e
is
gen
erall
y
co
ns
ti
tuted
by
th
e
relat
ion
:
n
=
d
+
p.
I
nd
i
vidu
al
par
it
y
bit
dr
ives
f
or
the
pa
ri
ty
of
sever
al
gr
oups
of
data
bits,
in
cl
ud
in
g
it
sel
f,
t
o
be
odd
(
or
e
ven),
w
he
re
e
ve
ry
pa
rity
is
ca
lc
ulate
d
on
di
f
fer
e
nt
s
ubset
s
of
the
data
bits.
T
he
bits
of
the
c
odewor
d
are
num
ber
e
d
s
uccessi
vely
,
be
ginnin
g
w
it
h
bit
1
at
the
le
ft
en
d,
bit
2
to
it
s
i
m
m
ediat
e
righ
t,
an
d
s
o
on.
In
Ham
m
ing
c
od
e
s,
the
par
it
y
bits
and
dat
a
bits
are
posi
ti
on
ed
at
a
part
ic
ular
place i
n
the codew
ord
. Th
e
pa
rity
b
it
s o
ccu
py posit
ion
s
2
0
, 2
1
, 2
2
. .
. 2
p
-
1
in the s
e
qu
e
nce
w
hic
h
has
at
m
os
t
2
p
-
1
−
1
posit
ions.
The
le
ft
ov
e
r
posit
ion
s
a
re
preserv
e
d
for
the
data
bits,
see
Figure
2
.
F
or
a
cod
e
w
ord
of
n
bit
s
,
there
are
2
n
po
ssible
cod
e w
ords
hav
i
ng
va
lues
from
0
to
2
n
−
1,
the
on
ly
2
d
of
them
are
valid
co
de
w
ords
an
d
2
n
–
2
d
are
no
n
-
valid
code
w
ords
.
3.
PREP
ROCES
SIN
G
OF
TE
ST D
ATA
Test
data
obta
ined
from
ATP
G
is
s
ubj
ect
e
d
to
prep
ro
ce
s
sing
ste
ps
li
ke
don’t
care
bit
fill
ing
a
nd
sp
li
tt
ing
of
test
d
at
a int
o
s
uitable
sli
ces.
3.1.
Sli
ci
ng
of
Inp
u
t
Te
s
t D
ata
We
div
i
de
the
input
te
st
dat
a
into
sca
n
c
ha
ins
of
predete
rm
ined
le
ng
t
h.
Let
us
ass
ume
that
the
te
st
data
T
D
con
sis
ts
of
n
te
st
patte
rn
s.
W
e
di
vi
de
the
scan
el
e
m
ents
into
m
scan
chai
ns
in
the
best
-
ba
la
nced
m
ann
er
possi
bl
e.
This
res
ult
in
eac
h
vect
or
bei
ng
d
i
vid
ed
int
o
m
s
ub
-
vectors,
ea
ch
of
le
ngth,
say
l
.
Dissim
il
arit
y
i
n
the
le
ngths
of
the
s
ub
-
vect
ors
are
reso
l
ved
by
pa
dd
i
ng
don’t
cares
at
the
en
d
of
the
sh
ort
e
r
su
b
-
vecto
rs.
T
hu
s
,
al
l
the
su
b
-
vect
or
s
are o
f
equ
al
le
ngth.
T
he
m
-
bit
data
wh
ic
h
is
pr
ese
nt
at
t
he
sa
m
e
po
sit
io
n
of
eac
h
s
ub
-
ve
ct
or
c
on
sti
tute
an
m
-
bit sli
ce.
If
the
re ar
e
vec
tors
at
the
beg
i
nn
i
ng, w
e
obta
in a total of
n × l
m
-
bit sli
ces, whic
h
is
our u
nco
m
pr
ess
ed
d
at
a se
t t
hat n
ee
ds
t
o be c
om
pr
essed
.
3.2.
D
on’t
’ C
are Bit
Fil
li
ng
The
te
st
c
ub
e
ge
ne
rated
by
a
uto
m
at
ic
te
st
patte
rn
gen
e
rato
r
(
ATPG
)
to
ol
con
ta
in
s
a
gr
e
a
t
qu
a
ntit
y
of
don’t
care(
X)
bits.
Su
c
h
don’t
care
bits
in
te
st
cub
e
can
be
m
anipu
la
te
d
for
e
nhanci
ng
t
he
te
st
data
com
pr
essio
n.
I
n
sta
ti
sti
cal
co
ding
te
ch
nique
s,
te
st
data
is
sp
li
t
into
e
qua
l
siz
e
sli
ces
of
m
bits.
Test
data
com
pr
essio
n
m
ay
be
im
pr
oved
by
re
duci
ng
the
num
ber
of
disti
nct
sli
ces
in
a
gi
ve
n
te
st
set
an
d
al
so
by
increasin
g
the
fr
e
qu
e
ncy of occ
urren
ce
for
e
ach d
ist
inct sl
ic
e. I
n
this
hybri
d
com
pr
essio
n schem
e, w
e app
ly
an
exi
sti
ng
don’t
care
bit
fill
ing
al
gorithm
na
m
el
y
MT
-
fill
wh
ic
h
has
le
ss
c
om
pu
ta
ti
on
al
c
om
plexity
co
m
par
e
d
to
oth
e
r
al
gorithm
s.
W
e
ha
ve
cho
se
n
Mi
ni
m
u
m
Tran
sit
ion
Fil
l
(MT
-
fill
)
over
oth
e
r
te
chn
i
qu
e
s
owin
g
to
th
e
fact t
hat it
r
e
du
ces the
num
ber
of wei
ghte
d
tr
a
ns
it
ion
s
in
t
he
test
v
ect
or, t
he
reb
y
reducin
g
t
he
te
st
power.
In
MT
-
fill
, a
pr
ogressi
on
of
X
en
trie
s in
the t
est
v
ect
or
is
fill
ed
with a
n
in
di
sti
ng
uis
hab
le
value
as t
he
first
non
-
X
ent
ry
on
t
he
rig
ht
side
of
this
ar
r
ang
em
ent.
Thi
s
lim
it
s
the
qu
a
ntit
y
of
tr
ansiti
on
s
i
n
the
te
st
vecto
r
wh
e
n
it
is sca
nned
in.
Fo
r
exam
ple,
consi
der
the
te
st
vecto
r:
100XX
010X
1X0.
This
vector,
af
te
r
MT
-
fill
,
w
ou
l
d
beco
m
e
100000
101100
.
If
t
he
te
st
vec
tor
has
a
stri
ng
of
X
bits
that
is
no
t
te
rm
inate
d
by
a
non
-
X
bit
on
the
ri
gh
t
sid
e
,
then
it
s
houl
d
be
fill
ed
by
t
he
bit
val
ue
to
the
le
ft
of
the
s
equ
e
nce.
F
or
e
xam
ple
10000010
11X
X
s
hould
be
100000
101111
after
MT
-
fill
.
4.
THE
PROPO
SED
METHO
DOLO
GY
We
pro
pose
an
i
m
ple
m
entation
of
a
hybri
d
com
pr
essio
n
schem
e
fo
r
red
uci
ng
the
vo
l
um
e
of
te
st
data.
O
ur
pro
po
s
ed
sc
hem
e
is
pri
m
aril
y
base
d
on
er
r
or
co
rr
ect
in
g
Ha
m
m
ing
co
des.
The
Ham
m
ing
c
ode
introd
uces
ad
di
ti
on
al
bits,
kn
own
as
pa
rity
bits,
w
ho
se
functi
on
is
to
va
li
date
the
exac
tness
of
the
or
iginal
m
essage
sent
upon
r
ecei
pt.
T
his
m
et
ho
d
tra
ns
f
or
m
s
the
sli
ce
of
siz
e
m
bi
ts
into
n
by
ad
ding
up
p
pa
rity
bits,
base
d
on
the
siz
e
of
the
m
essage
m
,
w
hich
is
encode
d
into
a
cod
e
w
ord
of
le
ng
th
n
.
Fi
gur
e
4
show
s
the
bl
oc
k
diag
ram
o
f
the
te
chn
iq
ue
4
.
1
.
Seq
uence
of Com
pressi
on
St
ep
s
Our pr
opose
d
appr
oach ta
kes
the
ATPG
ge
ne
rated
or
i
gin
al
te
st vector a
nd
i
m
ple
m
ents a no.
of step
s,
as foll
ows:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
9
33
–
9
40
936
St
ep
1:
Stac
kin
g o
f
t
he
te
st d
at
a file
for
e
nc
od
i
ng.
St
ep
2
:
Co
nv
e
r
sion
of
eac
h
m
-
bit
sli
ces
into
an
n
-
bit
le
ngth
wh
e
re
(m
>n)
by
app
ly
ing
t
he
Ham
m
ing
dec
od
e
rs
on
t
he
input
te
st
vector.
It
is
worth
m
entio
ni
ng
her
e
t
ha
t
ever
y
sli
ce
of
siz
e
m
bits
i
s
no
t
the
c
od
e
word;
so
m
e
non
-
c
ode
w
ord
m
-
bit
s
li
ces
m
ay
al
so
exist.
S
o,
an
e
xtra
bit
is
put
i
n
us
e
t
o
diff
e
r
entia
te
betwee
n
non
-
cod
e
w
ords
a
nd
co
de
wor
ds
.
W
e
ap
pe
nd
a
bit
‘1’
if
the
bit
sli
ce
is
a
cod
e
w
ord
a
nd
bit
‘0’
if
it
is
no
t
a
cod
e
w
ord
to
a
n
ad
diti
on
al
fi
le
cal
le
d
an
add
e
d
bit
file
.
The
te
st
data
sli
ce
that
is
a
val
id
co
dewo
rd
is
conve
rted
t
o
a
n
n
-
bit
sli
ce
instea
d
of
an
m
-
bit
sli
ce.
In
valid
c
od
e
w
ord
is
s
hifted
a
s
it
is,
with
out
any
com
pr
essio
n,
i
nto
t
he
c
om
pr
essed bina
ry f
il
e
.
St
ep
3:
F
ur
t
he
r
c
om
pr
ession
of test
v
ect
or
fi
le
b
y ap
plyi
ng
Huff
m
an
co
ding.
St
ep
4
:
R
un
Len
gth
E
nc
oding
(RLE
)
a
nd
the
Huff
m
an
encodin
g
al
gor
it
h
m
are
ap
plied
to
ad
de
d
bi
t
file
.
Af
te
r
wa
rd
the
file
is
ad
de
d
to
the
hea
der
of
t
he
c
om
pr
essed
bin
a
ry
file
.
As
a
res
ult
,
we
at
t
ai
n
a
pleasi
ng
r
esult
with
no
te
worth
y im
pr
ovem
ent
in
c
om
pr
essio
n rati
o.
Fig
u
re
3
si
gn
ifie
s
the
flo
w of
t
he pr
opose
d m
et
ho
d.
4
.
2
.
Dec
om
pr
ession
Mech
ani
sm
A
deco
m
pr
ess
ion
process
is
carrie
d
ou
t
i
n
order
t
o
reinstat
e
the
or
i
gin
al
te
st
dat
a
from
the
com
pr
essed
d
a
ta
f
il
e an
d
is
pe
rfor
m
ed
as
fo
ll
ow
s:
St
ep
1
:
Re
a
d
t
he
head
e
r
of
th
e
com
pr
esse
d
f
il
e
and
ta
ke
out
the
ad
de
d
bit
file
from
that,
after
that,
dec
ode
th
e
extracte
d
file
by
ap
plyi
ng
t
he
Huff
m
an
dec
od
e
r.
A
pp
ly
th
e
RLE
dec
od
e
r
in
order
t
o
re
instat
e
the
ad
de
d
bi
t
file
to
it
s origi
nal form
.
St
ep
2:
Decodi
ng of t
he
c
om
pr
esse
d data f
il
e
b
y m
eans of
t
he Huffm
an
al
gorithm
.
St
ep 3
:
E
ncodi
ng
o
f
the Huf
f
m
an
decode
d
te
st
data
file
by
app
ly
in
g
the H
am
m
ing
erro
r
co
rr
ect
io
n
al
gorithm
.
In
t
his
phase,
ever
y
bit
of
t
his
file
is
acce
ssed
an
d
th
us
t
he
le
ng
th
of
eve
ry
sli
ce
is
dep
e
nd
e
nt
on
t
he
val
ue
in
the
added
bit
fil
e.
If
the
bit
of
added
bit
fil
e
is
1,
then
the
slice
siz
e
is
n
bits,
otherwis
e,
the
siz
e
of
the
slice
is
m
bits.
The
Ham
m
ing
enc
od
e
r
returns
the
sli
c
e
of
siz
e
n
bits
to
it
s
or
iginal
siz
e
of
m
bits
and
retu
rn
s
t
he
par
it
y
bits that
wer
e
rem
ov
ed
du
rin
g Ham
m
ing
dec
od
i
ng.
St
ep
4
:
Last
ly
,
the test
data fi
le
is retu
rn
e
d
t
o
it
s
or
igi
nal st
at
us
without a
ny
loss of
data
Figure
3
.
Flo
w
of the
co
m
pr
e
ssion sch
em
e
Figure
4
.
Sc
he
m
e il
lustrati
on
us
in
g
t
he bloc
k
diag
ram
T
e
s
t
V
e
c
to
r
S
p
l
it
T
e
s
t
V
e
c
to
r
i
nto
S
l
ic
e
s
o
f
‘
m’ b
i
ts
A
p
p
l
y
Huff
man
Erro
r
C
o
rr
e
c
t
io
n
C
o
de
o
n
e
a
c
h sli
c
e
C
he
c
k
w
h
e
th
e
r
the
s
l
ic
e
is
a
vali
d
k
e
yw
o
rd
A
p
p
e
nd
‘
0’
to
th
e
add
e
d
b
it
f
il
e
Enc
o
d
e
th
e
s
l
ic
e
us
i
ng Ha
mm
in
g
Erro
r
C
o
rre
c
tin
g
c
o
d
e
s
D
e
c
o
de
th
e
s
l
i
c
e
us
in
g
d=
n
b
its
A
dd
1
to
th
e
a
dd
e
d
b
it
fil
e
C
o
p
y t
he
s
l
i
c
e
to
th
e
c
o
mp
r
e
s
s
e
d
fil
e
A
p
p
l
y
run
l
e
ngth
e
nc
o
di
ng
(R
LE)
o
n
the
ad
de
d
b
it
fi
l
e
A
p
p
l
y
Huff
man
e
nc
o
di
ng
al
g
o
rithm
o
n
t
he
add
e
d
b
i
t
fi
l
e
M
e
rg
e
add
e
d
b
i
t
fi
l
e
w
ith
the
h
e
ade
r
o
f
the
c
o
m
p
r
e
s
s
e
d
te
s
t
dat
a
fil
e
C
o
mp
re
s
s
e
d
T
e
s
t
D
a
ta
F
i
l
e
Huffma
n
al
g
o
ri
thm
fo
r
o
ut
p
ut
c
o
m
p
r
e
s
s
io
n
Y
e
s
No
0
0
1011010
0 0 0
1
010
0
0110
111
SoC
-
IP
C
o
r
e
Te
s
t V
e
cto
r
F
or
ma
t
i
o
n o
f
m
-
bi
t s
l
i
c
e
s
0
0
1011
00
00
0
10
1000
11
0
111
011
010
00
0
10
1000
111
H
am
min
g E
r
r
o
r
-
C
o
r
r
e
c
tin
g
C
o
d
e
s
C
o
de
s
C
o
de
s
C
omp
r
e
s
s
e
d
F
i
l
e
A
pp
l
i
cat
i
o
n o
f
H
u
f
f
ma
n
e
nc
od
i
ng
al
gor
i
th
m
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A SoC
-
IP
C
or
e
Test Dat
a
Co
mp
re
ssio
n Sc
he
me base
d o
n Error
Correcti
ng Ha
mm
i
ng Cod
e
s
(
Sanj
oy M
it
ra
)
937
5.
E
X
PERI
MEN
TAL
RESU
L
TS A
ND AN
A
LYSIS
In
order
t
o
ob
serv
e
t
he
li
kely
i
m
pr
ov
em
ents
to
be
delive
red
by
the
pr
opose
d
m
et
ho
d,
exp
e
rim
ents
are
car
ried
out
on
the
se
ven
IS
CA
S’89
[
20
]
ben
c
hm
ark
c
ircuit
s.
Sy
nopsy
s
Tet
ra
MA
X
[21]
ATPG
t
oo
l
is
us
e
d
to
ge
ner
a
te
the
te
st
data
.
Syn
opsy
s
Tet
ra
MA
X
was
f
un
ct
io
nal
with
the
dy
nam
ic
c
om
pacti
on
tu
rned
on
and
ra
ndom
-
fill
turn
e
d
off.
T
he
pr
opos
e
d
e
nc
od
i
ng
schem
e
base
d
on
the
Ham
m
ing
co
de
is
analy
zed
f
r
om
the
sever
al
po
i
nts
of
view
of
t
he
te
st
data
com
pr
essio
n
a
nd
t
he
ir
eff
ect
s
on
c
om
pr
essio
n:
t
he
com
pr
essio
n
rati
o,
siz
e o
f
sli
ce an
d
num
ber
o
f pa
rity
b
it
s etc.. Th
e ab
ove issue
s ar
e v
ery im
po
rta
nt in
the conte
xt of th
e pr
opos
e
d
schem
e. Th
e c
om
pr
essio
n
rat
io,
R
(
D
), t
hat
is estim
at
ed
w
it
h
the
fo
ll
owi
ng
100
D
E
D
D
R
T
T
T
T
C
(1)
Wh
e
re
|
D
|
den
otes
the
siz
e
of
the
or
i
gin
al
te
st
data
and
the
siz
e
of
the
com
pr
esse
d
te
st
data
is
rep
re
sented
by
|
E
|
Ta
bl
e
1
.
Som
e
D
et
a
il
s
of
Te
st
Da
ta
in
Singl
e
Sc
an
Chai
n
Archi
te
c
t
ure
ISCAS 89
circuit[
2
0
]
Total n
o
of
p
atterns
Bits
per pattern
Origin
al
Test
data
size(T
D
)
in
bits
s5
3
7
8
111
214
2
3
,75
4
s9
2
3
4
159
247
3
9
,27
3
s1
3
2
0
7
236
700
1
,65
,20
0
s1
5
8
5
0
126
611
7
6
,98
6
s3
5
9
3
2
16
1763
2
8
,20
8
s3
8
4
1
7
99
1664
1
,64
,73
6
s3
8
5
8
4
136
1464
1
,99
,10
4
Table
2
.
C
om
pr
essio
n
Ra
ti
o
ba
sed
C
om
p
arison w
it
h t
he
Pr
e
viously
Pro
pos
ed
Tec
hniq
ues
ISCAS 89
circuit[
2
0
]
Selective
Hu
f
f
m
an
[
4
]
Go
lo
m
b
[
7
]
FDR[
8
]
EFDR[
1
0
]
ALT
-
FDR[
1
1
]
RL
-
Hu
f
f
m
an
[
1
2
]
9
C[
2
2
]
Ou
r
Prop
o
sed
s5
3
7
8
4
2
.32
3
7
.11
4
8
.02
5
3
.67
4
5
.39
4
6
.17
5
1
.56
6
4
.60
s9
2
3
4
3
8
.14
4
5
.25
4
3
.59
4
8
.66
3
5
.32
4
2
.0
5
0
.91
6
1
.27
s1
3
2
0
7
6
6
.95
7
9
.74
6
9
.59
8
2
.19
2
9
.11
6
9
.51
7
2
.31
6
3
.65
s1
5
8
5
0
5
2
.61
6
2
.82
5
6
.82
6
7
.82
2
5
.90
5
7
.83
6
6
.37
7
2
.35
s3
5
9
3
2
5
0
.71
4
3
.21
4
4
.07
3
9
.41
3
4
.30
5
5
.08
5
7
.45
5
8
.30
s3
8
4
1
7
7
9
.87
2
8
.37
8
5
.17
6
2
.03
2
2
.41
8
9
.44
6
0
.41
6
8
.76
s3
8
5
8
4
5
7
.80
5
7
.17
6
0
.84
6
1
.12
2
3
.60
6
1
.52
6
5
.53
7
2
.60
In
orde
r
to
ve
rify
the
e
ff
i
ci
ency
of
the
pro
po
s
ed
m
et
hod,
im
ple
m
entat
ion
of
th
e
pro
posed
com
pr
essio
n
m
et
ho
d
was
ca
rr
ie
d
out
in
the
‘C’
la
ngua
ge
on
a
Lin
ux
sys
tem
.
Test
cub
e
s
we
re
gen
e
rated
f
or
the
se
ven
la
rgest
IS
CA
S
-
89[
16
]
fu
ll
y
sca
nned
te
st
be
nc
h
ci
rcu
it
s.
Tet
ra
Ma
x
[14]
A
T
PG
t
oo
l
was
use
d
t
o
gen
e
rate
these
te
st
data
cub
es
.
Table
1
giv
e
s
the
descr
ipti
on
of
the
sca
n
c
hain
net
wor
k
schem
e.
In
Tab
le
1,
we
prov
i
de
th
e
detai
ls
li
ke
nu
m
ber
of
pat
te
rn
s,
bits
per
patte
rn
an
d
t
otal
bits
of
te
st
data
cu
be
f
or
t
he
tradit
ion
al
si
ngle
scan
chai
n
a
rch
it
ect
ure
et
c.
The
the
num
ber
of
te
st
patte
rn
s
pro
vid
e
d
i
n
col
um
n
2,
m
ay
be
furthe
r
re
du
ce
d
by
ap
plyi
ng
com
pacti
on
.
I
f
the
com
pacte
d
te
st
vector
of
the
scan
c
hai
n
is
exactl
y
equ
al
to
oth
e
r
c
om
pacte
d
te
st
vectors, t
he
te
st
vecto
r
c
an be
rej
ect
e
d f
ro
m
the test
cube.
Table
3.
E
val
ua
ti
on
of
Com
pressi
on Rat
io
ba
sed o
n
Di
ff
e
re
nt Slic
e Sizes
a
nd
Va
riat
ion
in
the
Nu
m
ber
of
Parity
Bi
ts
Sl
i
ce Siz
e
(
S
)
ISCA
S
89
ci
rc
u
i
t
S
=
7
S
=
1
0
S
=
1
5
S
=
2
0
S
=
2
5
N
o
of
p
ar
i
t
y
bi
t
s
(
p
)
N
o
of
p
ar
i
t
y
bi
t
s
(
p
)
N
o
of
p
ar
i
t
y
bi
t
s
(
p
)
N
o
of
p
ar
i
t
y
bi
t
s
(
p
)
N
o
of
p
ar
i
t
y
bi
t
s
(
p
)
p
=
3
p
=
4
p
=
5
p
=
3
p
=
4
p
=
5
p
=
3
p
=
4
p
=
5
p
=
3
p
=
4
p
=
5
p
=
3
p
=
4
p
=
5
s
5
3
7
8
5
7
.
6
7
5
7
.
7
0
5
7
.
7
3
5
7
.
7
6
5
8
.
8
0
5
8
.
8
3
5
7
.
8
1
5
9
.
7
0
5
9
.
7
1
5
7
.
8
2
5
9
.
7
2
6
1
.
9
0
5
7
.
9
1
5
9
.
7
8
6
4
.
6
0
s
9
2
3
4
5
4
.
3
0
5
4
.
3
6
5
4
.
4
6
5
4
.
3
3
5
5
.
3
0
5
5
.
4
1
5
4
.
5
0
5
6
.
3
0
5
6
.
3
2
5
4
.
5
3
5
6
.
4
5
5
8
.
3
6
5
4
.
6
7
5
6
.
5
2
6
1
.
2
7
s
1
3
2
0
7
5
5
.
9
1
5
5
.
9
4
5
5
.
9
5
5
5
.
9
7
5
7
.
9
7
5
7
.
9
9
5
5
.
9
8
5
8
.
6
1
5
8
.
6
4
5
5
.
9
9
5
9
.
8
5
6
1
.
4
9
5
6
.
0
5
5
9
.
8
7
6
3
.
6
5
s
1
5
8
5
0
6
2
.
3
7
6
2
.
4
2
6
2
.
4
5
6
1
.
4
0
6
3
.
2
1
6
3
.
2
2
6
1
.
4
2
6
5
.
7
1
6
5
.
7
8
6
1
.
5
0
6
5
.
9
0
6
8
.
2
1
6
1
.
5
0
6
5
.
9
4
7
2
.
3
5
s
3
5
9
3
2
4
6
.
7
2
4
6
.
8
1
4
6
.
9
3
4
6
.
7
6
4
9
.
0
3
4
9
.
1
0
4
6
.
7
7
5
2
.
8
8
5
2
.
9
7
4
6
.
8
0
5
2
.
9
1
5
5
.
7
3
4
6
.
8
0
5
3
.
7
2
5
8
.
3
0
s
3
8
4
1
7
6
0
.
1
0
6
0
.
1
5
6
0
.
2
4
6
0
.
1
4
6
1
.
4
0
6
1
.
4
4
6
0
.
3
2
6
3
.
9
0
6
3
.
9
7
6
0
.
4
9
6
4
.
1
2
6
6
.
3
8
6
0
.
4
0
6
4
.
2
1
6
8
.
7
6
s
3
8
5
8
4
6
3
.
6
2
6
3
.
7
2
6
3
.
8
0
6
3
.
7
3
6
5
.
8
7
6
5
.
9
0
6
3
.
7
8
6
7
.
8
7
6
8
.
0
0
6
3
.
7
9
6
8
.
3
0
7
0
.
9
0
6
3
.
8
0
6
9
.
2
0
7
2
.
6
0
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
9
33
–
9
40
938
Table
2
pr
e
sen
ts
a
com
par
iso
n
of
the
c
om
pr
ession
rati
o
wi
th
sel
ect
ive
H
uffm
an
cod
i
ng
[
4],
G
olo
m
b
cod
i
ng
[
7],
F
DR
co
di
ng
[
8],
EF
DR
c
od
i
ng
[
10]
,
ALT
-
F
DR
Co
ding
[11],]
,
RL
-
Huffm
an
co
ding
[12],
a
nd
9C
[
22
]
.
F
or
e
ach
te
st
be
nch
ci
rcu
it
,
we
ha
ve
us
e
d
five
diff
e
ren
t
te
st
dat
a
sli
ce
siz
es
(i.e.
S=7
,
S=
10,
S=1
5,
S=2
0,
S=2
5)
and
thre
e
di
fferent
le
ng
t
hs
of
pa
rity
bits
(i.
e.
p=
3,
p=
4,
p=
5)
.
Col
um
n
9
giv
es
the
best
com
pr
essio
n
ra
ti
o
of
the p
r
op
os
e
d
com
pr
ess
ion
sc
hem
es
based
o
n
the n
um
ber
of
pa
rity
bi
ts
of
e
rro
r
co
rrec
ti
ng
Ham
m
ing
co
de
an
d
siz
e
of
t
he
te
st
data
sli
ce.
T
his
hy
br
i
d
c
om
pr
essio
n
pro
posal
outp
erfor
m
s
m
os
t
of
the
p
re
viously
publ
ished
c
om
pr
ession
sc
hem
es t
abu
la
te
d i
n
Ta
ble 2
for
the m
ajorit
y of
t
he
te
st ben
c
h
ci
rc
ui
ts. The
pro
po
se
d
sc
he
m
e
co
m
pr
esse
s
the
te
st
data
thrice
in
three
con
sec
utive
di
sti
nct
co
m
pr
e
ssion
m
et
ho
ds
:
firstl
y
with
Ham
m
ing
cod
e
ba
sed
c
om
pr
ession,
th
e
n
by
RLE
an
d
finall
y
with
t
he
ai
d
of
H
uff
m
an
cod
in
g
an
d
thus
yi
el
ds
i
m
pr
ess
ive
te
st
data
c
om
pr
essio
n
rat
io
in
m
os
t
of
t
he
ben
c
hm
ark
ci
rcu
it
s
us
ed
in
the
e
xp
e
rim
ent.
In
Table
2,
it
can
be
obse
rv
e
d
th
at
this
te
st
data
com
pr
ession
s
chem
e
has
resu
lt
ed
in
sig
n
ifi
cant
i
m
pr
ovem
ent
of
the co
m
pr
essi
on
rati
o
in
all
th
e li
ste
d
be
nc
hm
ark
circ
uits e
xcep
t t
he be
nc
hm
ark
s13
207 and s
38417.
In
Ta
ble
3,
it
can
be
seen
th
at
var
ia
ti
on
i
n
the
pa
ram
et
er
s
li
ke
sli
ce
siz
e(S)
a
nd
no
of
pa
rity
bits
durin
g
th
e
co
m
pr
ession
with
Ham
m
i
ng
er
ror
c
orrecti
ng
cod
e
s
is
in
flue
ncin
g
the
ove
r
al
l
com
pr
essio
n
rati
o
of
this
hybr
i
d
co
m
pr
ession
sch
e
m
e.
Her
e
in
Table
3,
it
is
e
vid
e
nt
that
in
m
os
t
of
the
cases,
with
the
increase
in
te
st data sli
ce
s
iz
e and
no o
f
pa
rity
b
it
s,
the c
om
pr
essio
n
rati
o
is al
so
i
m
pr
ov
e
d
in contras
t t
o
the p
rev
io
us on
e
.
If
we
crit
ic
al
ly
ob
se
r
ve
the
c
om
pr
ession
rati
o
im
pr
ov
em
ent
patte
rn
in
Tab
le
3,
we
ca
n
noti
ce
that
signi
ficant
i
m
pr
ovem
ent
i
s
rec
orded
in
case
of
{
(
S=7
,p
=
3)
t
o
(
S=1
0,p=4)},
{
(S
=
10,
p=4)
to
(S
=
15,
p=4)},{
(S
=
15,
p
=4
)
to (S=2
0,p=
5)
}
and
(S
=
20,p
=
5) to
(S
=
25,p
=
5)
}
for m
os
t of the
be
nch
m
ark
circ
uits
Table
4.
Var
ia
t
ion
of
Com
pr
e
ssion Ra
ti
o
with the
Pe
rcen
ta
ge
of the
Vali
d Sl
ic
e
,
Size
of
Sli
ce
and
Qu
a
ntit
y
of
Pa
rity
Bi
ts
ISCAS 89
circuit
Slice Size
(S),
No
o
f
parity
b
its
(
p
),
Pe
r
cent
ag
e of
the v
alid
slices
(V
S
)
S
=7
,
p
=3
S
=1
0
,
p
=4
S
=1
5
,p=
4
S
=2
0
,p=
5
S
=2
5
,p=
5
%
o
f
valid
slice
(V
S
)
CR
%
o
f
valid
slice
(V
S
)
CR
%
o
f
v
alid
slice
(V
S
)
CR
%
o
f
valid
slice
(V
S
)
CR
%
o
f
valid
slice
(V
S
)
CR
s5
3
7
8
25
5
7
.67
37
5
8
.80
39
5
9
.70
37
6
1
.90
35
6
4
.60
s9
2
3
4
28
5
4
.30
40
5
5
.30
41
5
6
.30
40
5
8
.36
32
6
1
.27
s1
3
2
0
7
23
5
5
.91
38
5
7
.97
31
5
8
.61
33
6
1
.49
34
6
3
.65
s1
5
8
5
0
34
6
2
.37
41
6
3
.21
37
6
5
.71
34
6
8
.21
30
7
2
.35
s3
5
9
3
2
27
4
6
.72
43
4
9
.03
45
5
2
.88
39
5
5
.73
37
5
8
.30
s3
8
4
1
7
32
6
0
.10
28
6
1
.40
41
6
3
.90
40
6
3
.38
38
6
8
.76
s3
8
5
8
4
34
6
3
.62
30
6
5
.87
35
6
7
.87
34
6
5
.90
31
7
2
.60
It
can
be
cl
earl
y
unde
rstoo
d
f
ro
m
F
igur
e
3
a
nd
F
ig
ure
4
t
ha
t
so
m
e
of
the
ham
m
ing
er
r
ors
co
rr
ect
in
g
cod
e
s
are
vali
d
a
nd
so
m
e
oth
ers
are
non
-
va
li
d.
I
n
Ta
ble
4,
we
ha
ve
s
how
n
t
he
per
ce
ntage
of
vali
d
sli
ce
s
capab
le
of
generati
ng
valid
c
od
e
s
w
hich
in
t
urn
pr
oduces
t
he
init
ia
l
le
vel
of
te
st
data
co
m
pr
ession.
In
Table
4,
the
pe
rce
ntage
of
valid
sli
ce
s
co
rr
es
pondin
g
to
var
io
us
te
st
data
sli
ces
are
show
n.
Her
e
the
valid
sli
ce
per
ce
ntage
V
S
corres
pondin
g
the
pair
(slic
e
siz
e,
no
of
pari
ty
bits)
agai
nst
dif
fer
e
nt
benchm
ark
te
st
da
ta
an
d
their
res
pecti
ve
com
pr
ession
rati
o
(CR)
is
al
so
giv
e
n.
High
e
st
com
pr
ession
rati
o,
72.
60
is
achie
ve
d
fo
r
t
he
pair
(S
=
25, p=
5) h
a
ving
31%
of v
al
id
sli
ces.
Ap
a
rt
from
the
ta
bu
la
r
re
pr
es
entat
ion
of
the
exp
erim
ental
data,
we
ha
ve
al
so
pu
t
the
gli
m
ps
e
of
the
exp
e
rim
ental
ou
tc
om
es
in
gr
a
ph
i
cal
re
presen
ta
ti
on
with
t
he
ai
d
of
a
col
umn
-
bar
c
ha
rt.
I
n
fig
ur
e
5,
be
nchm
ark
wise
com
par
ison
of
com
pr
es
sion
rati
o
(CR)
against
the
pa
ir
of
sli
ce
siz
e
and
no
of
pa
rity
bits
is
sh
own
.
The
pairs:
{(
S=
7,
p=3),
(S
=
10,
p=4),
(S
=
15,
p=4),
(S
=
20,
p=5)
an
d
(
S=
25,
p=
5)
}
a
nd
their
co
rr
es
pondi
ng
ben
c
hm
ark
cir
cuit wise c
om
pr
essio
n rati
os
a
re
plo
tt
ed
i
n
F
i
gure
5 for
gr
a
phic
al
stagin
g o
f
the
co
m
par
is
on
Com
par
ison
of
diff
e
ren
t
com
pr
essi
on
m
e
tho
dolo
gies
with
our
pro
posed
m
et
ho
d
on
the
basis
of
th
e
com
pr
essio
n
ra
ti
o
in
dif
f
ere
nt
ben
c
hm
ark
ci
r
cuits
is
plo
tt
ed
in
fig
ur
e
6.
T
hi
s
propose
d
m
e
thod
was
c
om
par
ed
with
oth
er
exi
sti
ng
m
et
ho
ds
nam
el
y
sel
ec
t
ive
H
uffm
an
cod
i
ng
[
4],
G
ol
om
b
cod
i
ng
[
7],
F
DR
co
ding
[
8]
,
EFD
R
c
o
di
ng
[10],
AL
T
-
F
D
R
Cod
in
g
[
11]
,
RL
-
H
uffm
an
cod
i
ng
[12],
a
nd
9C
[
22
]
and
alm
os
t
in
m
ajo
rity
of
the
ben
c
hm
ark
ci
rcu
it
s,
the
ba
r
corres
ponding
to
our
pro
pose
d
com
pr
ess
ion
sc
hem
e
is
sta
nd
i
ng
highe
st
with
highest c
om
pr
ession rati
o am
on
g al
l othe
r
c
om
pr
ession
m
eth
ods.
The
beh
a
vior
of
dif
fer
e
nt
be
nch
m
ark
ci
rc
ui
ts
in
te
rm
s
of
com
pr
essio
n
r
at
io(CR)
co
rre
sp
on
ding
to
the
pairs
of
sli
ce
siz
e
and
nu
m
ber
of
pa
rity
b
it
s
is
gr
ap
hi
cal
ly
dep
ic
te
d
in
F
ig
ur
e
7.
H
ere,
the
c
om
pr
ession
rati
o
corres
pond
i
ng
to
the
pa
irs:
{(
S=7
,
p=
3),
(S
=
1
0,
p=
4),
(S
=
15,
p=
4),
(S
=
20,
p=
5)
a
nd
(
S=2
5,
p=
5)}
against
the
be
nch
m
ark
ci
rc
ui
ts
us
ed
in
th
is
exp
e
rim
ent
are
plo
tt
ed
i
n
F
ig
ur
e
7.
F
ro
m
this
gr
ap
hical
pr
ese
ntati
on,
we
m
ay
cor
relat
e
how
the
c
om
pr
essio
n
rat
io
is
var
yi
ng
acro
s
s
dif
fer
e
nt
ben
c
hm
ark
ci
rcu
it
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
A SoC
-
IP
C
or
e
Test Dat
a
Co
mp
re
ssio
n Sc
he
me base
d o
n Error
Correcti
ng Ha
mm
i
ng Cod
e
s
(
Sanj
oy M
it
ra
)
939
d
epe
ndin
g
on
t
he
com
bin
at
io
n
of
the
siz
e
of
the
sli
ce
and
nu
m
ber
of
pa
ri
ty
bits
us
ed
in
the
Ham
m
ing
error
correct
ing c
odes.
Figure
5.
Ben
ch
m
ark
wise
comp
ari
son of
CR
bas
ed
on
the c
om
bi
nat
ion
of
sli
ce si
ze
(S) a
nd
num
be
r
of
p
ari
t
y
bit
s
(p
)
Figure
6
.
CR bas
ed
compari
son
of
diffe
r
ent c
om
pre
ss
ion
m
et
hod
s with
th
e
propo
sed
m
et
hod
on
d
iffe
ren
t
b
enc
hm
a
rk
ci
rcu
it
s
Figure
7.
CR b
ase
d
beh
avi
or
an
aly
sis of
different
benc
hm
ark
ci
rcu
it
s de
p
endi
ng
on
the c
om
bination
of
sli
ce siz
e
(S)
and
num
ber
of
par
ity bit
s (p)
6.
CONCL
US
I
O
N
This
pap
e
r
pr
e
sented
a
te
st
c
om
pr
essio
n
al
gorithm
that
com
bin
es
the
a
dv
a
ntage
s
of
t
he
Ham
m
ing
error
-
co
rr
ect
i
ng
c
od
es
,
RLE
an
d
Huff
m
an
e
ncodin
g.
T
his
pa
pe
r
devel
op
e
d
the
ef
f
ic
ie
nt
util
iz
at
i
on
of
Ham
m
ing
er
ror
-
c
orrecti
ng
co
des
in
c
om
bin
at
ion
wit
h
RL
E
an
d
H
uf
fm
a
n
enc
odin
g
al
gorithm
fo
r
te
st
data
com
pr
essio
n
in
orde
r
t
o
im
pr
ov
e
c
om
pr
ession
rati
o
of
SoC
-
IP
co
re
te
st
data.
We
ha
ve
app
li
ed
our
al
gorithm
on
va
rio
us
be
nch
m
ark
s
a
nd
com
par
ed
ou
r
res
ults
wit
h
existi
ng
te
st
c
om
pr
essio
n
te
chn
i
qu
e
s.
Our
hybri
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
9
33
–
9
40
940
com
pr
essio
n
s
chem
e
ou
tper
f
or
m
s
oth
er
e
xi
sti
ng
te
st
data
com
pr
essio
n
in
a
sig
nifica
nt
m
ann
er
,
gi
ving
a
best
po
s
sible
com
pr
essio
n
of
72.60%.
Si
gn
i
fican
t
i
m
pr
ov
em
ent
in
com
pr
essio
n
ef
fici
ency
is
ob
s
er
ved
at
th
e
cos
t
of
pro
ba
bly
li
ttle
increase
i
n
on
-
c
hip
dec
oder
area
ov
e
rh
e
ad.
Furthe
r
im
pro
vem
ent
of
com
pr
essio
n
r
at
io
a
nd
on
-
c
hip
dec
od
er
area
m
ini
m
iz
at
ion
for
suc
h
hybri
d
te
st
data
com
pr
ession
sc
hem
es
m
ay
be
the
f
uture
pros
pects
of
researc
h
in
this
par
ti
cular
sub
-
pro
blem
.
Sign
ific
an
t
featur
es
f
r
om
the
com
pr
essi
on
appr
oach
es
[23
-
25]
m
ay
al
s
o
be
inc
orpor
at
ed
in
ordr
t
o
fr
am
e
up
m
or
e
eff
ic
ie
nt
hybr
id
c
om
pr
essio
n
m
echan
ism
.
REFEREN
CE
S
[1]
P.T
.
Gon
c
ia
r
i
,
B
.
M.
Al
-
Hashim
i
and
N.
Ni
col
i
ci
,
“
Impr
ovi
ng
compress
ion
ratio,
a
rea
ove
rhead
an
d
te
st app
li
ca
ti
o
n
ti
me
for
system
-
on
-
a
-
chi
p
t
est
d
ata
compress
ion/
dec
ompr
ession
,
”
Proc.
Design
,
Autom
at
ion
and
Te
st
in
Europe
Conf.
,
2002.
[2]
Z
You
,
W
W
ang
,
Z
Dou
,
P
L
iu
a
nd
J
Kuang
“
A s
ca
n
disab
li
ng
-
ba
sed
BAS
T
sche
m
e
for
te
st c
ost r
educ
t
ion,
”
IEI
C
E
El
e
ct
ron.
Ex
pres
s
,
vol. 8
(2011) pp 1367
-
1373
[3]
Jas
,
J.
Ghos
h
-
Dastida
r
,
Mom
-
Eng
Ng
and
N.A.
Toub
a
“
An
eff
icient
te
st
vec
tor
compress
ion
sche
m
e
using
sele
c
ti
ve
Huffm
an
codi
ng
,
”IEE
E
Tr
ansacti
on
on.
Computer
-
Ai
d
ed
Design
Inte
gr
ati
on
Circui
ts
Sy
stem
s,
vol.
22(6)
2003,
pp
797
-
80
6
[4]
X
Kavousia
nos
,
E
Kall
ig
ero
s
and
D
Nikolos
“
Opt
imal
sele
c
ti
v
e
Huffm
an
codi
ng
for
te
st
-
data
compress
ion,
”
I
EEE
Tr
ansacti
ons on Com
pute
rs
,
vol.
56
(8)
Jul
y
2007
,
pp
1146
-
115
2
[5]
P.
Sis
m
anogl
ou
and
D.
Nikolos
“
Input
te
st
da
ta
c
om
pre
ss
ion
base
d
on
the
reu
se
of
par
ts
of
dic
t
ionar
y
en
tries:
Stat
i
c
and
d
y
namic
ap
proa
che
s,
”
I
EEE
Tr
ansacti
ons
on
Computer
-
Aided
Design
o
f
I
nte
grated
Circu
it
s
and
S
yste
ms
,
vol.
32
(11)
O
ctober
2013,
pp
17
62
-
1775
[6]
T.
B.
W
u,
H
Z
Li
u
and
P
X
Li
u
“
Eff
ic
ie
nt
t
est
c
om
pre
ss
ion
te
chni
que
for
SO
C
b
ase
d
on
bloc
k
me
rging
and
ei
gh
t
codi
ng,
”
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
o
n
i
c
T
e
s
t
i
n
g
,
Vol.
29
(
6)
Dec
ember
20
13,
pp
849
-
859
[7]
Chandra
and
K.
Chakra
bar
t
y
“
Sy
stem
-
on
-
a
-
chi
p
te
st
-
da
ta
compr
ession
and
dec
o
m
pre
ss
ion
arc
hitect
ur
es
base
d
on
Golom
b
c
odes,
”
IEE
E
Tr
ansacti
ons
on
Co
mputer
-
Ai
ded
Design
of
Inte
grated
Ci
rcuit
s
and
Syste
m
s
,
vol.
20
(3)
Marc
h
2001
,
pp
355
-
368.
[8]
Chandra
and
K.
Chakra
bar
t
y
“
Te
st
data
compress
ion
and
te
st
resourc
e
par
t
it
io
ning
for
sy
st
em
-
on
-
a
-
chi
p
usin
g
fre
quency
dir
ec
t
ed
run
-
l
engt
h
(FD
R)
code
s,
”
I
EEE
Tr
ansacti
ons
on
Computers
,
v
ol
.
52
(8)
2003,
p
p
1076
-
1088
[9]
P.T
.
Gon
ci
ar
i,
B.
M.
Al
-
Hashim
i
,
and
N.
Nico
lici
,
“
Vari
able
-
l
en
gth
input Huffm
an
cod
ing
for
s
y
stem
-
on
-
a
-
chi
p
te
st”
IEEE Trans
ac
t
ions o
n
Com
pute
r
-
Aided
Desi
gn
of
Int
egr
a
te
d
Circ
uit
s
and
S
y
s
te
m
s
,
vol
.
22(6),
pp
783
–
796
,
June
2003
[10]
H.
El
-
Ma
le
h
“
Tes
t
data com
pre
ss
ion
for
s
y
stem
-
on
-
a
-
chi
p
using
e
xte
nded
fre
qu
en
c
y
-
d
irecte
d
run
-
l
engt
h
code,”
IET
Computers and D
igi
tal
Te
chni
qu
es
,
vol
.
2
(3)
Apr
il
2008
,
pp
155
-
163
[11]
Chandra
and
K.
Chakra
bar
t
y
“
A
unifi
ed
appr
o
ac
h
to
r
educe
SO
C
te
st
d
at
a
volume
,
sca
n
power
an
d
te
st
ing
t
ime,
”
IEE
E
Tr
ansacti
o
ns on
Co
mputer
-
Ai
ded
Design
of
Inte
grated
Circu
it
s and
Syste
ms
,
vol.
22
(3)
Marc
h
2003,
pp
352
-
363
[12]
M.
Nourani
and
M.
H.
Te
hra
ni
pour
“
RL
-
Huff
m
an
enc
oding
for
te
st
com
pre
ss
ion
and
power
red
uct
ion
in
sc
a
n
appl
i
ca
t
ion,”
AC
M
Tr
ansacti
ons
on
Design
Au
to
mation
of
Elec
tronic
S
yste
ms
,
V
ol.
10
(1)
2005,
p
p
91
-
115
[13]
Us
ha
S.
Mehta
,
Kanka
r
S.
Das
gupta
and
Nira
nja
n
M.
Deva
shra
y
ee
“
Run
-
le
n
gth
-
base
d
te
st
d
at
a
compress
io
n
te
chn
ique
s: How
far
from
ent
rop
y
and
pow
er
bou
nds
?
—
A surve
y
,
”
VLSI
D
esign
F
eb
2010
pp
1
-
9
[14]
R.
W
.
Ham
m
ing,
“
Err
or
detec
t
ing
and
e
rror
cor
r
ecting
cod
es,
”
The
Be
l
l
Syst
em
Tec
hnic
al
Journal
,
vol.
29(
2
)
April
1950,
pp
147
–
16
0.
[15]
Ta
nenb
aum,
Co
m
pute
r
Network
s,
Prentice
Hall, 2003.
[16]
G.
Cai
re
,
S.
S
hamai
and
S.
Verdu
“
Lo
ss
le
ss
data
compress
ion
wit
h
error
correc
ti
on
cod
e
s,”
Proc.
IEE
E
Inte
rna
ti
ona
l
S
ym
posium
on
Inform
at
ion
Th
eor
y,
2003
,
pp
22
[17]
G.
Cai
re
,
S.
Sham
ai
,
S.
Verdu,
“
A
new
data
compress
ion
algo
rithm
for
sour
c
es
wit
h
memor
y
based
on
erro
r
correc
ti
ng
cod
es,
”
Proc
.
IEEE
W
or
kshop on
Infor
m
at
ion
Th
eor
y
,
2003,
pp
291
–
29
5
[18]
A.A.
Sharie
h,
“
An
enha
nce
m
en
t
of
Huffm
an
c
oding
for
the
co
m
pre
ss
ion
of
mul
ti
m
edi
a
f
il
es,
”
Tr
ansacti
ons
o
n
Engi
ne
ering,
C
omputing
and
Te
chnol
ogy
,
Vol
.
3
2004
,
pp
303
–
3
05.
[19]
T.
C.
Be
ll
,
I.
H
.
W
it
te
n,
J.
G
,
Cl
e
ary
Text Compr
e
ss
ion
,
Prentice
Hall,
1990
.
[20]
F.
Brgle
z
,
D.
Br
y
an
and
K.
Kozm
inski
“
Co
mbinati
onal
pro
fi
le
s
of
seque
nt
i
al
benc
hmar
k
c
ircui
ts
,
”
In
IE
E
E
Inte
rna
ti
ona
l
S
ym
posium
on
Circ
uit
s
and
S
y
s
te
m
s,
Vol.
3
Ma
y
1
989,
pp
1929
-
1
934
[21]
S
y
nops
y
s In
c.: T
et
ra
MA
X ATPG
user
Guide
,
2
006.
[22]
M
Te
hra
nipoor,
M
Nourani
and
K
Chakra
bar
t
y
,
“
Nine
-
code
d
c
om
pre
ss
ion
te
chni
que
for
te
sting
embedde
d
cor
e
s
in
SoC
s”
.
IE
EE
Tra
nsac
ti
ons on
VLSI
Syste
ms
,
Vol.
13(6)
2005
,
pp
719
–
731
[23]
S
J.
Sarka
r,
N
K
Sarka
r,
T
Du
tt
a
,
P
De
y
,
and
A
Mukherje
e5
,
“
Arithmeti
c
Co
ding
Based
App
roa
ch
for
Pow
er
S
y
stem Para
m
eter
Data
Com
pre
s
sion”
,
Indone
sia
n
Journal
of
Ele
ct
rical
Engi
n
ee
ri
ng
and
Computer
Sci
ence
,
vol
.
2
(
2),
pp
.
268
-
274
,
Ma
y
2016
[24]
T
S
Gunawan,
M
Kart
iwi,
“
Perform
anc
e
Ev
al
u
at
ion
of
Mult
ic
h
anne
l
Audio
Co
m
pre
ss
ion”
,
Indone
sian
Journal
of
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Computer
Sc
ie
nc
e,
V
ol.
10(1)
,
pp
.
14
6
-
153,
April
201
8
[25]
W
Song,
“
Strat
egi
es
and
Te
chn
i
ques
for
Data
Com
pre
ss
ion
in
W
ire
le
ss
Sensor
Networks”
TEL
KOMNIKA
,
vol.11
(11),
pp
.
6624
-
6
630,
Novem
ber
2013
.
Evaluation Warning : The document was created with Spire.PDF for Python.