Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
5, N
o
. 4
,
A
ugu
st
2015
, pp
. 71
4
~
71
9
I
S
SN
: 208
8-8
7
0
8
7
14
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
A Scalab
le L
a
rge Form
at
Disp
lay Based on Zero Client
Processor
San
g
Don
Kim, Seu
n
g Eun
Lee
Department o
f
Electronic Eng
i
neering,
Seoul National Univ
ersity
of Scien
c
e and
Technolog
y
,
R
e
pu
blic of Korea
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Feb 10, 2015
Rev
i
sed
May 12
, 20
15
Accepted
May 30, 2015
This paper prop
oses the zero client m
odule that
targets Lar
g
e Format Display
(LFD) sy
stem for display
wall. I
n
creas
ed r
e
solution in modern L
F
D requires
a high bandwid
th chann
e
l and
a high perfor
m
ance display
controller
to
transfer th
e im
a
g
e dat
a
to th
e
m
onito
r. The k
e
y id
ea is to use
a Gigabit-
Ethern
et communication based
Dais
y
-
Chai
n to
trans
f
er an
im
ag
e dat
a
.
This
communication
supports sufficient bandwid
th
fo
r image data
transfer. As a
result, we implement the LFD sy
stem
using th
e
zero client module and
LCD
m
onitors.
Keyword:
FPGA
Gi
ga
bi
t
-
Et
he
rn
et
Large
F
o
rm
at Display
(
L
F
D
)
Zero Clien
t
Copyright ©
201
5 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Seun
g Eu
n Lee,
Depa
rt
m
e
nt
of
El
ect
roni
c
En
g
i
neeri
n
g,
Seou
l Nation
a
l
Un
iv
ersity o
f
Scien
ce an
d Tech
no
log
y
,
21
9 C
h
a
n
g
h
a
k
Hal
l
,
Se
oul
Na
t
i
onal
U
n
i
v
ersi
t
y
of
Sci
e
nce a
n
d
Tec
h
n
o
l
o
gy
, 2
3
2
Go
n
gne
u
n
g
gi
l
,
N
o
w
o
n-
gu
,
Seou
l 139
-74
3
, Repu
b
lic
o
f
Ko
r
e
a.
Em
a
il: seu
n
g
.lee@seou
ltech.ac.kr
1.
INTRODUCTION
The si
ze
of
s
i
ngl
e LC
D sc
reen
has c
o
nt
i
n
u
o
u
s
l
y
i
n
cre
a
sed acc
od
ri
n
g
t
o
t
h
e
deve
l
opm
ent
o
f
sem
i
cond
uct
o
r
p
r
oces
s.
The
l
a
rge
r
LC
D sc
reen
requires
t
h
e La
rge Fo
r
m
at D
i
sp
lay
(
L
FD
) wh
ich co
nn
ects
th
e
m
u
ltip
le LCD screen
s to i
m
p
l
e
m
en
t a
l
a
rg
e
d
i
sp
lay fo
r an
adv
e
rtisemen
t su
ch
as
b
illb
o
a
rd
s. The LFD
syste
m
req
u
i
res th
e
h
i
gh p
e
rfo
r
m
a
n
ce d
i
sp
l
a
y con
t
ro
ller to
d
r
i
v
e the m
u
ltip
le screen
s.
Howev
e
r, th
is
is v
e
ry
expe
nsi
v
e bec
a
use t
h
e di
spl
a
y
cont
rol
l
e
r i
n
cl
u
d
es t
h
e hi
gh
per
f
o
r
m
a
nce grap
hi
c car
d
for di
s
p
l
a
y
i
n
g
vi
a
HDM
I or VGA interface. These tra
d
itiona
l interfaces
ha
ve bee
n
restric
t
ed to tra
n
sm
is
sion
distance i
n
doze
n
o
f
me
t
e
r
s
.
I
n
ca
s
e
o
f
l
a
r
g
e
d
i
s
p
l
a
y
o
v
e
r
ma
x
i
mu
m
t
r
a
n
sm
iss
i
on
distance, the re
peater
shou
ld
b
e
ad
op
ted, and
it lead
s to
in
crease
o
f
t
h
e co
st fo
r th
e LFD system
. In
o
r
d
e
r to
sub
s
titu
te th
e trad
itio
n
a
l
d
i
sp
lay cab
l
e, a
Gi
ga
bi
t
-
Et
he
rn
et
can
be a
d
o
p
t
e
d.
The
Gi
gabi
t
-
Et
her
n
et
s
u
p
p
o
r
t
s
l
o
n
g
er
di
st
ance t
o
t
h
e
d
a
t
a
t
r
ansm
i
ssi
on, a
n
d
it h
a
s su
fficient b
a
ndwid
th for th
e LFD.
The LF
D sy
st
em
has bee
n
s
t
udi
ed
usi
ng
v
a
ri
o
u
s arc
h
i
t
ect
ure. R
u
d
o
l
f
s
et
al
. pr
op
ose
d
vi
de
o wal
l
cont
rolled by
m
i
ni PC [1].
Kim
et
al. propos
ed the c
ontrolling m
e
t
hod for
t
h
e vide
o wall
base
d on
the
network [2]. A central c
ont
rol ha
s sim
p
le
architectur
e
,
and it does not use a separat
e
controller for each
di
spl
a
y
.
On t
h
e i
nde
pen
d
e
n
t
cont
rol
,
eac
h
d
i
spl
a
y
m
odul
e
cont
rol
s
t
h
e vi
deo
si
g
n
al
p
r
o
v
i
d
e
d
f
r
om
t
h
e su
ppl
y
u
n
it.
Sub
a
sh
et
al. pro
p
o
s
ed
t
h
e
Gig
a
b
it-Ethern
et
b
a
sed
d
a
ta
acq
u
i
sition
syste
m
fo
r im
a
g
ing
array
[3
]. High
resol
u
tion im
a
g
e for a single
photh
on
X-ra
y syste
m
was success
f
ully tran
sm
itted through Gi
gabit-E
h
ternet
[4]
.
In th
is
p
a
p
e
r,
we
p
r
op
o
s
e the scalab
le LFD arch
it
ecture
using t
h
e client syste
m
whic
h incl
udes the
Giga
bit-Ethe
rnet interface. T
h
e re
st
of
our pa
per is
orga
nized as
fo
llows.
Section 2 desc
ribes a
bout the
back
g
r
o
u
nd
of
cl
i
e
nt
sy
st
em
s and L
F
D
.
A
n
d we
bri
e
fl
y
i
n
t
r
od
uces t
h
e
f
eat
ures
of t
h
e LFD sy
st
em
and t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 4
,
Au
gu
st 2
015
:
71
4
–
71
9
7
15
zero cl
i
e
nt
m
o
dul
es i
n
sect
i
o
n 3, a
nd
prese
n
t
t
h
e har
d
war
e
im
pl
em
ent
a
tion
of
ou
r p
r
ot
ot
y
p
e i
n
Sect
i
o
n
4
.
Sect
i
on
5 s
h
o
w
s t
h
e
ex
pe
ri
m
e
nt
al
resul
t
and
Sect
i
o
n
6
concl
ude
s t
h
i
s
pa
per
by
out
l
i
ni
ng
t
h
e
di
rec
t
i
on
fo
r
fu
t
u
re wo
rk
on
th
is
top
i
c.
2.
CLIENT SY
S
TEMS AN
D LFD
The Gi
g
a
bi
t
-
Et
her
n
et
base
d screen i
m
age t
r
ansm
i
ssi
on ha
ve bee
n
use
d
on t
h
e cl
i
e
nt
sy
st
em
s. The
client syste
m
,
whic
h is a kind of the rem
o
te access
term
inal, recei
ves image data to
cont
rol LC
D m
onitor
t
h
r
o
u
g
h
t
h
e Gi
gabi
t
-
Et
her
n
et
and t
r
a
n
sm
i
t
s
the use
r’s i
n
p
u
t
such as key
b
o
a
rd
or m
ouse.
In
or
der t
o
s
u
p
p
o
r
t
pers
o
n
al
des
k
t
op e
n
vi
ro
nm
ent
,
Vi
rt
ual
De
skt
o
p I
n
f
r
ast
r
uctu
re (VDI) allo
ws to
g
a
in
access to
th
e desk
top
envi
ro
nm
ent
thr
o
ug
h
t
h
e Gi
gabi
t
-
Et
her
n
et
com
m
uni
cat
i
on. It
su
pp
o
r
t
s
si
ngl
e or
m
u
l
t
i
pl
e
scree
n
out
put
fo
r
con
v
e
n
i
e
nce
.
The cl
i
e
nt
sy
st
em
depe
nds
o
n
a ser
v
e
r
t
o
s
u
p
p
o
rt
hi
g
h
pe
rf
orm
a
nce co
m
put
at
i
on. Thi
s
cl
i
e
nt
syste
m
is d
i
v
i
ded
i
n
to
t
h
ree categ
o
ries: a
fat clien
t
, a
t
h
i
n
c
l
i
e
nt
an
d a
zer
o cl
i
e
nt
.
Am
on
g t
h
ese sy
st
em
s, t
h
e
fat
cl
i
e
nt
has a hi
gh
per
f
o
r
m
a
nce p
r
oces
so
r
and o
p
erat
i
n
g
sy
st
em
. Even t
h
o
u
g
h
t
h
i
s
cl
i
e
nt
sy
st
em
provi
des
ri
ch
f
unct
i
o
nal
i
t
y
and i
t
ca
n
pr
ocess
m
a
ny
pr
o
g
ram
s
i
n
a
m
u
l
t
i
t
a
ski
n
g
envi
ro
nm
ent
,
i
t
co
nsum
es m
u
ch
o
f
po
we
r. O
n
t
h
e
ot
he
r ha
n
d
, T
h
e t
h
i
n
cl
i
e
nt
co
nsi
s
t
s
o
f
t
h
e l
o
w pe
rf
orm
a
nce
pr
ocess
o
r a
n
d ope
rat
i
n
g sy
st
em
. It
has no suffici
ent perform
a
nce for
progra
m which requi
res high computation powe
r. Because the
serve
r
sup
p
o
rt
s
su
ffi
c
i
ent
pr
ocessi
n
g
po
wer
t
o
e
x
ecut
e
m
o
st
of
a use
r’s
p
r
og
ram
,
t
h
e per
f
o
r
m
a
nce of t
h
e
cl
i
e
nt
syste
m
is n
o
t
sign
ifican
t.
Also
, It
h
a
s a lo
w
p
r
ice
and
low power
p
r
op
erties.
Howev
e
r, th
e t
h
i
n
clien
t
consum
es
m
u
ch of
powe
r than the ze
ro client system
. Th
e zero
clien
t
has th
e sim
p
lest
h
a
rdware
wh
ich
is
o
p
tim
ized
fo
r
a rem
o
te clien
t
syste
m
. On
th
e zero
clien
t
, th
e op
erating
syste
m
is n
o
t
essen
tial b
ecause th
e
zero clien
t
syste
m
estab
lish
e
s th
e
VDI ch
an
n
e
l
u
s
i
n
g th
e G
i
g
a
b
it-
Et
h
e
rn
et to supp
or
t
th
e per
s
o
n
al desk
t
op
envi
ro
nm
ent
.
A zer
o cl
i
e
nt
sy
st
em
depen
d
s o
n
t
h
e se
r
v
er t
o
s
u
p
p
o
r
t
com
put
at
i
on
po
we
r an
d ha
rd
war
e
resources, a
n
d
it im
ple
m
ent the sim
ilar user interface
ba
se
d
on the
com
putational
powe
r
of t
h
e se
rver. Mos
t
of t
h
e
program exec
ution is
proces
sed
on the server, and t
h
e client
receives the
proces
sing
res
u
lt as a
scre
e
n
im
age. In terms of power c
onsum
ption, the
zero client is
proper t
o
the
LFD system
because the
fat client and
th
e th
in clien
t
sp
en
t m
u
ch
of
po
wer t
h
an th
e zero clien
t
.
The
Giga
bit-Et
hernet ba
sed
L
F
D system
is s
i
m
i
lar to
zero
client syste
m
because the
scre
en im
age is
fed t
o
LC
D s
c
reen t
h
r
o
ug
h
Gi
ga
bi
t
-
Et
he
r
n
et
.
Ho
we
ver,
t
h
e Gi
gabi
t
-
Et
her
n
et
t
r
a
n
s
m
i
ssi
on su
ffe
r
s
t
h
e
b
o
ttlen
e
ck
when
it driv
es m
u
ltip
le LCD screen
s on
t
h
e
LFD system
. Th
erefo
r
e, th
e i
m
ag
e co
m
p
ression
alg
o
rith
m
is u
s
ed
on
t
h
e VDI. In
ord
e
r to
avo
i
d
t
h
e bo
ttleneck
o
n
t
h
e Gi
gab
it-Eth
e
rn
et ch
ann
e
l,
we ad
op
t th
e
Daisy-ch
ai
n
co
nn
ection
.
The Daisy-ch
ain is a wiri
n
g
sch
e
m
e
in
wh
i
c
h
m
u
ltip
le d
e
v
i
ces are con
n
ected
to
g
e
th
er i
n
series. It redu
ces th
e risk
o
f
a
b
o
ttlen
e
ck
o
n
th
e Gi
g
a
b
it-Et
n
e
rh
et chann
e
l
b
ecau
s
e each serial
eth
e
rn
et con
n
e
ctio
n
estab
lishes th
e lo
cal n
e
twork. Th
is
D
a
i
s
y
-
chai
n
bas
e
d o
n
t
h
e
Gi
g
a
bi
t
-
Et
he
rnet
c
a
n b
e
u
s
ed
for th
e LFD system
b
e
cau
se LFD syste
m
co
n
s
ists o
f
th
e m
u
ltip
le LCD screen
an
d
t
h
e clien
t
syste
m
w
h
ich
is con
t
ro
ls th
e scr
een. V
e
ndo
r
s
h
a
v
e
d
e
v
e
l
o
p
e
d
a zer
o
-
c
lien
t
b
a
sed
d
i
sp
lay, wh
i
c
h
suppo
r
t
str
e
a
m
lin
e
busi
n
ess IT e
n
vi
r
onm
ent
wi
t
h
V
D
I. T
h
e t
h
i
n
cl
i
e
nt
can be
used f
o
r t
h
e s
ecure V
D
I
en
v
i
ro
nm
ent
[5]
.
In or
de
r
to
establish se
cure V
D
I
,
the PC-o
ve
r-
IP (P
coI
P
) pr
oces
s
o
r ba
sed ze
r
o
cl
i
e
nt
sy
st
em
[6
]
and t
h
e zer
o
cl
i
e
nt
u
s
ing
th
e
PCoIP ch
ip
set, supp
orting
VM
ware [7
] were
d
e
v
e
lop
e
d. Th
e clo
u
d
conn
ect is si
m
ilar with
th
ese
clien
t
syste
m
s an
d it h
a
s po
rtab
ility [8
]. An
arch
itectur
e for
zero clien
t
b
a
sed
LFD system was
propo
sed
[9
].
In
our system
,
each LCD sc
re
en of the LF
D
system
has a single ze
ro client
m
odule and e
ach m
odule
is co
m
b
in
ed
with
Daisy-chain
b
a
sed
on
th
e Gig
a
b
it-Et
hernet. The proposed L
F
D
syste
m
has sc
alable
p
r
op
erties th
ank
s
to th
e
feature th
at
the LCD
scree
n
ca
n be attached or
de
t
ach
ed.
And
t
h
is is con
t
ro
lled
b
y
th
e
host
com
put
er
suc
h
as
pe
rso
n
a
l
com
put
er
vi
a Gi
ga
bi
t
-
Et
he
rnet
.
3.
Z
E
RO CLIENT
MODULE
FO
R SCAL
ABLE LFD SYSTEM
M
o
st
exi
s
t
i
ng
LFD co
nsi
s
t
s
o
f
a di
spl
a
y
part
and a di
spl
a
y
cont
rol
l
e
r (S
ee
Fi
gure
1(a
)).
The di
s
p
l
a
y
p
a
rt in
clud
es
m
u
l
tip
le LCD m
o
n
ito
rs, and
th
e d
i
sp
lay
co
nt
r
o
l
l
e
r su
pp
o
r
t
s
num
erous m
oni
t
o
r o
u
t
p
ut
p
o
rt
s i
n
or
der t
o
c
o
nt
r
o
l
t
h
e di
spl
a
y
part
. T
h
ere
f
ore
,
i
t
requi
res a
hi
g
h
com
put
at
i
on
po
we
r f
o
r
vi
de
o p
r
oce
ssi
ng
. To
red
u
ce t
h
e pe
r
f
o
r
m
a
nce requ
i
r
em
ent
,
we adopt
e
d
t
h
e Dai
s
y
-
C
h
ai
n co
n
n
e
c
t
i
on t
o
o
u
r p
r
op
ose
d
LF
D sy
st
e
m
(See Figu
re
1
(b)). An
i
m
ag
e d
i
sp
lay co
n
t
roller, wh
ich
is o
n
th
e
d
i
sp
lay co
n
t
ro
ller
p
a
rt, sp
lits th
e i
m
a
g
e fo
r
di
spl
a
y
pa
rt
, a
nd t
r
an
sm
it
s t
h
e di
vi
ded i
m
age fra
gm
ent
wi
t
h
a
n
I
D
.
O
n
t
h
e di
spl
a
y
part
,
a si
ngl
e ze
ro
c
l
i
e
nt
m
odul
e co
nt
rol
s
a di
s
p
l
a
y
m
oni
t
o
r
,
an
d eac
h
of t
h
em
i
s
con
n
ect
ed i
n
Dai
s
y
-
C
h
ai
n
usi
n
g
t
h
e Gi
gabi
t
-
Et
her
n
et
.
Each ze
ro clie
nt m
odule has
a uni
que m
o
dule ID on
the
LFD system
, and it enab
les c
l
assification of each
m
onitor. The i
m
age data can
be dis
p
layed when t
h
e received fragm
e
nt ID
is
m
a
tched with the m
odule ID of
zero client m
odule.
Othe
rwis
e, receive
d dat
a
is delivere
d
to the next m
odule through
Da
isy-Chain. T
h
e
r
efore
,
t
h
e im
age di
sp
l
a
y
cont
r
o
l
l
e
r
doe
s n
o
t
re
qui
re hi
gh
per
f
o
r
m
ance, an
d Gi
gabi
t
-
Et
her
n
et
con
n
ect
i
o
n s
u
pp
o
r
t
s
suf
f
i
c
i
e
nt
c
o
m
m
uni
cat
i
on ba
nd
wi
dt
h
fo
r L
F
D sy
st
em
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A
Sc
al
a
b
l
e
L
a
r
g
e
F
o
r
m
at
Di
s
p
l
a
y Base
d o
n
Zero
C
l
i
e
nt
Pr
ocess
o
r
(S
eu
ng Eun
Lee)
71
6
(a)
(b
)
Fi
gu
re
1.
Zer
o
C
l
i
e
nt
M
o
d
u
l
e
fo
r Scal
a
b
l
e
L
f
d Sy
stem
; (a).
Struct
ure
of the existing LFD syste
m
,
(
b
)
.
Stru
ctu
r
e of
th
e pr
opo
sed
LFD
system
3.1.
Little
Core for Z
ero Cl
ient System
The Ni
os II c
o
re is a s
o
ft
core
whic
h is
synt
hesizable
fo
r FP
GA
. T
h
is
core is optimized for
perform
a
nce according t
o
va
rious c
o
nf
igurations in the
LFD
syste
m
. One
of the
types
of
Nios II core
i
n
accorda
n
ce with confi
g
urations is a Nios
II/s core which i
s
a standard co
re to rem
ove a significant tra
d
e-off
in s
o
ftwa
re
pe
rform
a
nce. This
core s
u
pports
the ethe
rnet
IP
f
o
r
Altera
FP
GA
and has
s
u
fficient
performance
to
m
a
n
a
g
e
th
e
flow of im
ag
e d
a
ta.
We establish
two
Gi
gab
i
t
-
Et
her
n
et
cha
nnel
o
n
t
h
e FP
GA beca
use of
t
h
e
Dai
s
y
-
C
h
ai
n re
qui
res t
w
o c
o
m
m
uni
cat
i
on c
h
an
nel
s
.
In order to process a large num
b
er of the image data, available
m
e
m
o
r
y
space should be large to
sto
r
e im
ag
e d
a
ta in
m
e
m
o
ry. We ad
op
t th
e
SDRAM as the syste
m
m
e
mo
ry th
at sup
ports sufficien
t
me
m
o
ry
space for m
a
gangi
ng t
h
e image data. T
h
e
SDRAM m
e
mory
controller can be
gene
rat
e
d by m
eans of Qsys
t
ool
an
d i
s
expl
oi
t
e
d t
o
ut
i
l
i
ze t
h
e SDR
A
M
i
n
ou
r s
y
st
em
. Howe
v
e
r, cl
oc
k sy
nc
hr
o
n
i
zat
i
on p
r
obl
em
s
according to the I/O tim
i
ng of FPGA occ
u
r betwee
n
the SDRAM c
ontroller and th
e external SDRAM (See
Fig
u
re
2
)
. To
reso
lv
e t
h
is prob
lem
,
th
e SDR
A
M con
t
ro
ller
and
the Nios II/s
core oper
ate with diffe
rent clock,
whi
c
h
has a
fas
t
er p
h
ase a
b
out
1.
5
n
s t
h
an
t
h
at
o
f
t
h
e
ext
e
r
n
al
SDR
A
M
,
ge
n
e
rat
e
d
by
t
h
e
P
LL.
Fi
gu
re
2.
B
l
oc
k
di
ag
ram
of S
D
R
A
M
c
o
nt
r
o
l
l
e
r
3.
2.
Gi
g
a
bi
t-
E
t
herne
t
f
o
r
Dai
s
y-
Ch
ai
n
The
T
r
i
p
l
e
Sp
eed
Et
her
n
et
M
A
C
(T
SE
M
A
C
)
IP
i
s
p
r
o
v
i
d
e
d
fo
r deve
l
ope
r
w
h
o desi
gns
ha
r
d
wa
re
wi
t
h
t
h
e et
her
n
et
com
m
uni
cat
i
on, an
d a
n
d
i
t
supp
ort
s
1
0
M
bps,
10
0 M
bps a
n
d 1 G
b
ps spee
d o
n
et
her
n
et
co
nn
ection
.
In o
u
r LFD syste
m
, th
e au
to
-neg
o
tiation
fu
n
c
t
i
on o
f
Gi
g
a
b
i
t
-
Et
her
n
et
sel
ect
s ope
rat
i
on
spee
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 4
,
Au
gu
st 2
015
:
71
4
–
71
9
7
17
a
m
ong the
three ethernet s
p
eeds. T
h
e
Nios II/s c
o
re
co
n
t
rol
s
t
h
e
ope
ra
t
i
on
of
s
u
b
bl
ock
s
s
u
ch
as
t
h
e T
S
E
MAC core, the Scatter-Gather Direct Mem
o
ry A
ccess
(SGDM
A
) TX, and the SGDMA RX by
setting
regi
st
ers
of ea
ch IP (
S
ee Fi
g
u
re
3). T
o
s
u
p
p
o
r
t
hi
g
h
-
s
pee
d
dat
a
t
r
an
sfe
r
bet
w
ee
n a
m
e
m
o
ry
and t
h
e
TS
E
MAC core, the
SGDMA t
r
ansfers
data from syste
m
bus in
terconnect t
o
s
t
reaming
interface,
and vice versa
.
In
our system
, two SGDMAs, whic
h im
ple
m
ent the
role
of transm
itter and
receiver, a
r
e c
o
nnected to t
h
e
TSE
MAC core
on
the side of the
stream
ing interface to eliminat
e congestion of system bus
caused
by m
a
ssive
dat
a
o
n
t
h
e et
h
e
rnet
c
o
m
m
uni
cat
i
on.
The
de
scri
pt
o
r
m
e
m
o
ry
i
n
cl
u
d
es a
s
e
ri
es o
f
desc
ri
pt
o
r
s,
w
h
i
c
h i
n
vol
ve
inform
ation about the data to be transfe
rre
d. Because
the out
put cloc
k generate
d from
the oscillator of the
FPG
A i
s
n
o
t
an i
d
eal
si
g
n
al
,
t
h
e D
D
I
O
(
D
o
ubl
e
Dat
a
rat
e
In
p
u
t
Out
put
) m
odul
e i
s
used
fo
r creat
i
n
g ac
curat
e
edge
-al
i
g
ned
t
r
ansm
i
ssi
on cl
o
c
k
of
t
h
e e
x
t
e
r
n
al
P
H
Y
chi
p
.
Fi
gu
re
3.
B
l
oc
k
di
ag
ram
of T
r
i
p
l
e
S
p
ee
d Et
her
n
et
(
T
SE
)
h
a
rd
ware
3
.
3
.
VGA Cont
ro
ller
The VGA controller
visualize
s
the receive
d
data to
the m
onitor. The
resoluti
on of our
prot
otype
is
6
4
0
x
480
and
it su
pp
orts 24
b
it tru
e
co
lor.
Th
e
VGA
c
ont
rol
l
e
r
has a
n
i
m
age bu
ffe
r that stores im
age fram
e
to
d
i
sp
lay.
W
e
ad
op
t th
e SRAM b
u
f
fer to
opti
m
ize th
e sy
ste
m
perform
ance, and it ena
b
les fast im
age display.
Fig
u
re
4
illu
strates th
e
b
l
o
c
k
d
i
agram
o
f
th
e
VGA con
t
ro
ller.
Fi
gu
re
4.
B
l
oc
k
di
ag
ram
of V
G
A
co
nt
r
o
l
l
e
r
Because the monitor dis
p
lay ope
ration
has t
h
e hi
ghest
pri
o
rity in this bl
ock, pi
xel buffer DM
A rea
d
s
data
from
the SRAM as soon as possible. T
h
e
Nios II/s core
stores an im
age data to
th
e SRAM wh
en
th
e me
m
o
ry
is av
ailab
l
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A
Sc
al
a
b
l
e
L
a
r
g
e
F
o
r
m
at
Di
s
p
l
a
y Base
d o
n
Zero
C
l
i
e
nt
Pr
ocess
o
r
(S
eu
ng Eun
Lee)
71
8
4.
IMPLEME
N
TATION
We ad
o
p
t
t
h
e
Al
t
e
ra DE
2
-
1
1
5
b
o
a
r
d
[1
0]
f
o
r i
m
pl
em
ent
a
t
i
on o
f
t
h
e ze
r
o
cl
i
e
nt
m
odul
e and t
h
e
pr
o
pose
d
LF
D
sy
st
em
whose
feat
ure
s
i
n
cl
u
d
e
Al
t
e
ra C
y
cl
one I
V
FP
GA
, 1
28M
B
S
D
R
A
M
,
4M
B
SR
A
M
, and
two Ethe
rnet P
HY c
h
ips.
Our syste
m
includes a Nios II/s core
, an Av
alon syste
m
bus, an SDR
A
M
m
e
m
o
ry
cont
rol
l
e
r,
a t
w
o
Gi
gabi
t
-
Et
her
n
et
i
n
t
e
rfac
e
s, a
VG
A
co
nt
r
o
l
l
e
r an
d
a
JTA
G
deb
u
ggi
ng
ha
r
d
wa
re.
Ni
os
II/
s
core is provi
de
d by Altera
, and it supports 2 GB of
e
x
terna
l
m
e
m
o
ry address space. T
h
e
SDRAM c
ont
roller
is co
n
t
ro
lled
by th
e co
r
e
, and th
e 12
8 MB of
SD
R
A
M is
a
v
ai
l
a
bl
e. T
w
o
Gi
ga
bi
t
-
Et
he
rn
et
i
s
used
fo
r
Dai
s
y
-
C
h
ai
n c
o
n
n
ect
i
on
wi
t
h
ot
he
r
m
odul
es. It
p
r
ovi
des t
h
e s
u
ff
i
c
i
e
nt
ban
d
w
i
d
t
h
f
o
r i
m
age d
a
t
a
t
r
ansm
i
ssi
on. T
h
e
VGA con
t
ro
ller in
teg
r
ates an
SRAM co
n
t
ro
l
l
er that
used for a
n
im
age buffer.
In
o
r
de
r t
o
c
o
n
s
t
r
uct
t
h
e LF
D
sy
st
em
usi
n
g
a zer
o cl
i
e
nt
m
o
d
u
l
e
,
com
m
uni
cat
i
o
n
pac
k
e
t
sh
oul
d
be
con
f
i
g
ure
d
f
o
r
im
age dat
a
t
r
ansm
i
ssi
on. In t
h
e o
u
r i
m
pl
em
ent
a
t
i
on,
we u
s
e t
h
e UD
P pr
ot
oc
ol
beca
use
of i
t
s
sim
p
l
i
c
i
t
y
.
Thi
s
pr
ot
oc
ol
has
a sm
al
l packe
t
header c
o
m
p
are t
o
t
h
e TC
P/
IP p
r
ot
oc
ol
. Al
t
h
o
u
gh t
h
i
s
packe
t
doe
s
not
s
u
pp
o
r
t
bi
di
rect
i
o
nal
err
o
r c
o
r
r
ect
i
o
n
algorith
m
,
it
is no
t critical in
th
is system
.
5.
E
X
PERI
MEN
T
AL RES
U
L
T
S
In
or
de
r t
o
veri
fy
t
h
e w
h
ol
e sy
st
em
, one zer
o cl
i
e
nt
m
odul
e i
s
used
fo
r t
h
e im
age di
spl
a
y
cont
r
o
l
l
e
r.
This m
odule operates at 100 MHz an
d it process the im
age data for the LF
D system
. In our
veri
fication
envi
ro
nm
ent
,
t
h
i
s
ze
ro
cl
i
e
nt
m
odul
e
reads
t
h
e
6
4
0
⨉
480 of
im
ag
e d
a
t
a
fr
o
m
th
e f
i
l
e
system
o
f
the ho
st
com
puter a
n
d
splits i
m
age on each
fra
gm
ent to c
onst
r
uc
t
a
UDP packet.
The UDP
p
acket
, which contains the
im
age fragm
ent
an
d t
h
e ne
cessary
p
r
ot
oc
ol
hea
d
ers
,
i
s
t
r
ansm
i
t
t
e
d t
o
t
h
e di
s
p
l
a
y
part
t
h
ro
u
gh
Gi
ga
bi
t
-
Ethernet.
The
di
spl
a
y
p
a
rt
co
nsi
s
t
s
of
t
h
e t
w
o ze
ro
cl
i
e
nt
m
odul
es an
d t
h
e
LC
D m
oni
t
o
rs
. E
ach m
odul
e
cont
ai
n
s
t
h
e 1
28M
B
o
f
SDR
A
M
and t
h
e
4
M
B
of SR
AM
,
and t
h
ese o
p
e
r
at
e at
100M
H
z
. The SR
AM
i
s
used
for th
e im
ag
e
d
i
sp
lay bu
ffer
an
d
it supp
orts u
p
to
10
24
⨉
7
68
res
o
l
u
t
i
o
n.
On t
h
e di
s
p
l
a
y
part
,
one
of t
h
e zero
client m
odule
has m
o
dule ID 0,
a
n
d a
not
her one
has
a
m
o
dule
ID
1
(Se
e
Figure 5). The
receive
d UDP packet
contains a
pa
rt of the
dis
p
lay im
age, a
n
d the pa
rt of t
h
e i
m
age i
s
2
0
pi
x
e
l
s
i
n
ou
r
pr
ot
ot
y
p
e.
The
zer
o cl
i
e
nt
m
odul
e di
s
p
l
a
y
s
t
h
e rec
e
i
v
e
d
i
m
age dat
a
t
h
r
o
ug
h
Dai
s
y
-
C
h
ai
n
w
h
en
t
h
e fra
gm
ent
ID
m
a
t
c
hes wi
t
h
m
odul
e
ID.
Ot
he
r
w
i
s
e,
t
h
e im
age dat
a
pass t
o
t
h
e
next
m
odul
e
usi
n
g Dai
s
y
-
C
h
ai
n t
h
ro
u
gh t
h
e Gi
gabi
t
-
Et
h
e
rnet
.
Fin
a
lly, th
e 60
H
z
of
640
⨉
48
0 res
o
l
u
t
i
o
n
of t
h
e i
m
age has bee
n
pri
n
t
e
d o
n
t
h
e t
w
o LC
D scre
ens;
t
h
e
ex
tend
ed
d
i
sp
l
a
y reso
lu
tion
is 1
280
⨉
480
. In
add
itio
n, th
e d
i
sp
lay p
a
rt can
b
e
ex
tend
ed to
m
u
ltip
le
mo
n
itors
because the
connection
st
ructure
has expa
ndability.
Fi
gu
re
5.
P
hot
og
ra
ph
o
f
ou
r e
xpe
ri
m
e
nt
al
envi
r
onm
ent
6.
CO
NCL
USI
O
N
In t
h
i
s
pape
r,
we p
r
esent
e
d
ou
r LF
D sy
st
em
whi
c
h co
ns
i
s
t
s
of t
h
e zer
o cl
i
e
nt
m
odu
l
e
and t
h
e
m
o
n
ito
rs. Th
e
h
o
s
t m
o
du
le wh
ich
is on
the d
i
sp
lay con
t
ro
ller sid
e
, sp
lits i
m
ag
e an
d
tran
sm
its UDP pack
et
through Giga
bit-Ethernet. T
h
e display pa
rt consists
of
m
u
ltiple zero client
m
odules
and each
of
th
em
con
n
ect
ed
i
n
D
a
i
s
y
-
C
h
ai
n. T
h
ese zero cl
i
e
nt
m
odul
e has a uni
q
u
e m
odul
e ID fo
r cl
assi
fi
cat
i
on of t
h
e m
odul
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 4
,
Au
gu
st 2
015
:
71
4
–
71
9
7
19
Each zero client m
odule cont
rols the m
onitor for the
im
ag
e display using the receive
d UDP pac
k
et from
the
host
m
odul
e.
Th
e exp
e
rim
e
n
t
al resu
lt d
e
m
o
n
s
trates th
e feasib
ility o
f
ou
r
propo
sal to
th
e LFD
syste
m
. Th
e
i
m
p
l
e
m
en
ted
syste
m
h
a
s an
ex
p
a
nd
ab
ility th
an
ks to
th
e con
n
ection
topo
lo
g
y
, and
it can co
n
s
titu
te th
e
m
o
re
l
a
rge di
spl
a
y
s
y
st
em
.
Al
t
hou
gh
t
h
e di
spl
a
y
cont
rol
l
e
r pr
oc
esses
t
h
e 64
0
⨉
48
0 i
m
age bec
a
use t
h
e c
o
m
put
at
i
o
n
po
we
r
of z
e
r
o
cl
i
e
nt
sy
st
em
is rest
ri
ct
e
d
, t
h
i
s
res
o
l
u
t
i
o
n c
o
ul
d
be
i
m
prove
d acc
or
di
n
g
t
o
t
h
e p
r
ocessi
n
g
po
w
e
r
o
f
th
e
pro
cesso
r.
On th
e
d
i
sp
lay p
a
rt, th
e
SRAM im
ag
e b
u
ffer is in
ad
eq
u
a
te
fo
r a
h
i
g
h
reso
lu
ti
o
n
imag
e
because the siz
e
. The
r
efore, t
h
is im
age buffer can
be
re
placed wit
h
a
high
er ca
pacity me
m
o
ry suc
h
as
DDR
m
e
m
o
ry. It is helpful for the high res
o
l
u
tion i
m
age displa
y on each ze
ro client m
odule.
In the fut
u
re
work, we
plan t
o
im
prove the pe
rform
a
nce
of t
h
e zero client sy
ste
m
and optimize the
UDP pac
k
et
to im
ple
m
ent
a m
o
re
efficient syste
m
.
ACKNOWLE
DGE
M
ENTS
Thi
s
st
udy
wa
s
sup
p
o
r
t
e
d
by
t
h
e IT R
&
D
pr
og
ram
of M
S
IP/
IIT
P [1
0
0
4
7
0
8
8
,
Devel
o
p
m
ent
of ope
n
t
y
pe Ha
do
o
p
st
ora
g
e a
p
pl
i
a
nc
e t
o
s
u
pp
ort
m
o
re
t
h
a
n
48
TB
pe
r si
n
g
l
e
dat
a
n
ode]
.
REFERE
NC
ES
[1]
Rudolfs Bundulis
, e
t
.al
.
,
“
C
oncept
of Virtu
a
l
M
achin
e bas
e
d High Resolution Display
W
a
ll”,
In
formatio
n,
Electronic and
Electrical Engin
e
ering,
pp
. 1-6
,
N
ov 2014.
[2]
Kwang-Yong Kim
, et.a
l.
, “The
Design of Server Virtualizati
on
Based Video-wall Control S
y
stem”,
Internationa
l
Conference on
C
o
mputing and C
onvergence Tech
nology,
pp. 401-
404, Dec 2012
.
[3]
Subash K
, et
.al
.
, “Design and
d
e
velopment of
a
Giga-bit Ethern
et b
a
sed high
s
p
eed bro
a
dband
data acquisitio
n
s
y
s
t
em
for
an un
derwater
im
agin
g arra
y”
,
OCEANS 2010,
pp
. 1
-
7
,
Sept. 2010
.
[4]
Sang Don Kim
, et.a
l.
, “Multi-En
erg
y
X-Ray
Im
aging S
y
st
em
Using Single Photo
n
Counting”,
In
ternational Review
on Computers a
nd Software,
vo
l. 8
,
no
. 7
,
July
, 2
013.
[5]
Samsung, Zero
& Th
in Clients
,
http
://www.sams
ung.com/us
/business/computing/zero-clien
t
s-and-
thin-clien
ts.
[6]
Terad
i
ci, Simplif
y
and Save with
PCoI
P Zero Clients: Down Size the End Po
int, h
ttp://www.terad
i
ci.com/products-
and-solutions/pcoip-products/zer
o-clients
[7]
LG,
LG
Zero Client
& Teradici P
c
oIP,
http:/
/www.
lg.
c
om/us/ commercia
l/desktop-virtualization
[8]
Dell, Cloud
Client-Computing an
d De
sktop Virtu
a
lization, http
://www.dell.
com/us/business/p/clo
ud-connect/pd
[9]
Sang Don Kim
, et.a
l.
, "Archit
ect
ure for A Zero
C
lient
Bas
e
d S
c
al
able
Larg
e F
o
rm
at Dis
p
la
y S
y
s
t
e
m
,"
Internationa
l
Journal of Computer and
In
formation Technolog
y,
2015
.
[10]
Altera,
DE2-115 Development and Educ
ation Board,
http://www.
altera.
c
om
/education/univ/mate
rials/boards/de2-
115/unv-de2-11
5-board.html
BIOGRAP
HI
ES OF
AUTH
ORS
Sang Do
n Kim
is M.S. student in the Depar
t
ment of
Electron
ic Engineeri
ng at
Seoul National
University
o
f
Scien
ce and Technolog
y
,
Seoul,
Korea. He has a B.S. degree in Electronic
Engineering fro
m Seoul National
University
of
Science
and T
echnolog
y
,
Seo
u
l, Korea. His
res
earch
in
teres
t
includ
es
com
put
er ar
chi
t
ec
ture
,
l
o
w power c
i
rcui
t
des
i
gn
and s
i
gn
al pro
ces
s
i
ng.
Se
ung Eu
n Le
e
is
an As
s
i
s
t
ant P
r
ofes
s
o
r in the Dept
. of El
ectron
i
c Eng
i
ne
ering at S
e
ou
l
National Univer
sity
of Scien
c
e and Technolog
y
.
He has a P
h
.D. degr
ee in
Electrical
and
Com
puter Engin
eering
from
the
Univers
i
t
y
of C
a
liforn
i
a,
Irvin
e.
His
curren
t
res
earch
int
e
res
t
s
includ
e com
pute
r
archi
t
ec
ture
,
m
u
lti-processor
s
y
stem
-on-chip
, low-power
and resili
ent
VLSI,
and hardware acceler
a
tion for emerging applicati
ons. Prior to joining Seoul Tech in 2010, he
s
p
ent 2
years
a
s
a platform
ar
chit
ect
at Int
e
l
Labs
, Hil
l
os
boro, OR, US
A and 6
years
as
a
res
earch
er at
Korea El
ectron
i
cs
Techno
log
y
Ins
t
i
t
ute
,
Kore
a.
Evaluation Warning : The document was created with Spire.PDF for Python.