Anda di halaman 1dari 15

31.1 / S. S.

Kim

31.1: Invited Paper: Worlds First 240Hz TFT-LCD Technology


for Full-HD LCD-TV and Its Application to 3D Display
Sang Soo Kim*, Bong Hyun You, Heejin Choi, Brian H. Berkeley*,
Dong Gyu Kim, and Nam Deog Kim
LCD Business, Samsung Electronics Co., Ltd.
Asan-Si, Chungcheongnam-Do, Korea
*

Technology Center, Samsung Mobile Display Co., Ltd.


Yongin-Si, Gyeonggi-Do, Korea

Abstract
A full-HD (FHD) liquid crystal display TV (LCD-TV) has been
enhanced by increasing the panels frame rate to 240Hz, and this
240Hz driving technology has been applied to 3D TV. Compared
to a 120Hz LCD, the 240Hz LCD has two challenges: 1) half of
the available pixel charging time, and 2) three times as many
interpolated frames. A new architecture has doubled the available
pixel charging time by means of a half-gate two-data (hG-2D)
driving scheme and a charge-shared super PVA (CS S-PVA) pixel
structure. Additionally, a 240Hz ME/MC algorithm has been
implemented on the LCD module to convert 60Hz incoming
frames into 240Hz frames. Motion picture response time (MPRT)
of the new LCD-TV has been measured as 4.7msec, which is
similar to the MPRT for a cathode ray tube (CRT) TV.

1.

charging time for HD, FHD, and UD resolutions driven at 60Hz,


120Hz, and 240Hz frame rates. Figure 2 shows that charging time
for FHD at 240Hz would only be 3.7s for conventional driving
methods. This time is insufficient for charging of a-Si thin film
transistors (TFTs) on a large area panel. Moreover, the S-PVA
pixel structure for wide viewing angle also needs two different
data values (low and high gamma) for a single pixel. Therefore, to
drive the LCD panel at 240Hz while maintaining a single bank of
column drivers, available charging time must be doubled.

1/120sec

Interpolated
frame

Introduction

Motion blur in LCDs mainly arises from the LCDs inherent hold
type driving method, and is considered to be one of the most
significant causes of moving picture image degradation. Recently,
due to rapid advances in image processing and LCD driving
technology, double speed (120Hz/100Hz) driving with motion
estimation (ME) and motion compensation (MC) has became
popular as a means of improving motion image quality [1].
However, the MPRT of 120Hz LCDs is on the order of 7-8ms,
which is similar to that of plasma display panels (PDPs). Motion
image quality of LCDs has still been considered to be inferior to
other display devices, especially CRTs. As a result, the need for
high speed driving beyond 120Hz has been noted. In this work,
we report the worlds first 240Hz LCD-TV panel, which has been
developed by adopting a half-gate two-data (hG-2D) and chargeshared super PVA (CS S-PVA) pixel structure and advanced
driving scheme [2-3].
The concept of 240Hz ME/MC driving is shown and compared to
conventional 120Hz ME/MC driving in figure 1. With 240Hz
driving, three interpolated frames are inserted between the two
original 60Hz frames. The extra frames result in hold time which
is half that of a 120Hz-driven LCD, which should result in further
reduction of motion blur. However, it is not enough to simply
send twice as many frames to the panel. To benefit from this extra
data, the LCD module must be completely re-architected from its
input interface to the pixel structure.

2.

Pixel Structure and Driving Method

2.1

hG-2D CS S-PVA pixel structure

To achieve 240Hz driving of a Super-PVA panel, the first point


that must be considered is to find a suitable countermeasure for
insufficient pixel charging time. Figure 2 shows a graph of pixel

424 SID 09 DIGEST

Original
Frames
(a) 120Hz frame rate driving

1/240sec

Interpolated
frames

Original
Frames
(b) 240Hz frame rate driving
Figure 1. Comparison of 120Hz ME/MC and 240Hz ME/MC
To meet this need, the 240Hz LCD panel incorporates hG-2D
driving and a CS S-PVA pixel structure to double the available
charging time without loss of angle of view. The principle of CS
S-PVA pixel driving is shown in figure 3. In the S-PVA structure,
one pixel is composed of low gamma (brighter) and high gamma
(darker) sub-pixels to widen the viewing angle. As a result, the SPVA structure requires two separate data for the two sub-pixels.
The CS S-PVA pixel structure has solved this problem by
adopting a charge sharing scheme via the Cdown capacitance. In
figure 3, one pixel data is charged simultaneously to sub-pixel A

ISSN/009-0966X/09/3901-0424-$1.00 2009 SID

31.1 / S. S. Kim
(low gamma, high brightness) and sub-pixel B (high gamma, low
brightness) while the gate-on signal GN turns on TFT1 and TFT2.
Then, the next lines gate signal (GN+1) turns on the next lines
TFTs, but also TFT3 to move a portion of the charge from subpixel B to Cdown. As a result, sub-pixel B becomes darker than
sub-pixel A due to the shared (reduced) charge into Cdown, and SPVA driving is achieved with only a single data value.

instead of using the gate-on signal of next line (GN+1). The gate
driver supplies CS control signals through additional output
channels as shown in fig.4. Therefore, this new method enables
simpler panel design which does not require a bridge line between
gate lines.

GN

Available Charging Time (s)

25

GN+1

20

120Hz UD
(hG-2D)

FHD
14.8

H
L

H
L

H
L

H
L

H
L

H
L

H
L

H
L

H
L

H
L

CSN+1

15

UD
7.4

GN+2

240Hz FHD
(hG-2D)

10.4

CSN+2

7.4

H
L

CSN

HD
20.7

10

H
L

3.7

60Hz
(16.7ms)

GN+3
4.9

2x

3.7
2.5
180Hz
(5.5ms)

120Hz
(8.3ms)

2x

1.8
240Hz
(4.1ms)

CSN+3

Figure 4. CS S-PVA panel structure with hG-2D driving and


separate CS control

Frame Rate (frame time)

Figure 2. Pixel charging time of HD, FHD, and UD resolutions


at various frame rates

Data

to TFT3 of GN-1
Clc(A)

Cst(A)

Sub-pixel A

2.2

240Hz ME/MC and panel driving scheme

To drive a panel at 240Hz, three frames must be interpolated for


every incoming 60Hz frame. The 240Hz frame stream consists of
double the data of 120Hz driving; therefore a more advanced
interface must be used between the timing controller (TCON) and
LCD panel. The new FHD 240Hz driving scheme has evolved in
two steps. Figure 5 shows the first step, which uses two frame rate
converters (FRCs) and two TCONs, each of which were originally
optimized for FHD 120Hz driving.

TFT1
VA
GN
VB
TFT2
Clc(B)

TFT3
Cst(B)

Sub-pixel B
Cdown

GN+1
Figure 3. Principle of CS S-PVA scheme
Use of hG-2D to drive one pixel with a single data value has
increased available charging time for one pixel from 3.7s to
7.4s, which is same as the time for 1G-2D FHD 120Hz as shown
in figure 2. In the hG-2D driving scheme, two gate lines are
turned on at the same time, and pixel data on each line is provided
by two separate data lines. Therefore, it is possible to charge all
pixels of two gate lines simultaneously and the charging time for
each line is doubled (3.7s 7.4s). Although the CS S-PVA
and hG-2D methods were introduced in our 82 UD display [4], in
the proposed 240Hz driving scheme, a more advanced topology
uses a separate charge-sharing (CS) signal to control CS timing

Figure 5. 240Hz ME/MC driving architecture (step 1)

SID 09 DIGEST 425

31.1 / S. S. Kim
In this structure, the two FRCs are dedicated to the left and right
halves of the FHD screen. The pixel frequency of FHD 120Hz is
the same as half-FHD 240Hz, and the re-programmed FRCs are
therefore capable of generating half-FHD data streams at 240Hz.
Because each FRC provides an interpolated stream for only half
of the screen, care must be taken to avoid artifacts near the panels
centerline. To prevent such artifacts, motion vector information
from the right side of the screen is needed by the left FRC engine,
and vice versa. In fig. 5, each FRC engine receives the full screen
incoming 60Hz FHD data stream by way of an LVDS repeater,
which duplicates the entire input signal. Then the two FRCs
generate interpolated frames for the left (columns 1-960) and right
(columns 961-1920) halves of the screen, transmitting the results
to each TCON using an LVDS interface. Each TCON sends its
data to the panels column drivers by way of an advanced intrapanel interface (AiPi) [5]. AiPi is a point to point interface which
uses clock-embedded differential signaling. It enables the pixel
data to be transmitted at 800Mbps, which reduces the number of
required signal lines.

and TCON have the same amount of input/output pixel data as in


step 1 (two FRCs and TCONs), changing the interface is essential
to reasonably manage the number of I/O pins and the package
size. With those improvements, the 240Hz driving scheme of step
2 is simpler, with less logic devices and lower ASIC count.
Table 1. Technology comparison for external(TCON input) and
internal (TCON output to LCD driver) interfaces

Number of signal lines


between FRC and TCON

Conventional
Technology
LVDS
96 lines

Advanced
Technology
V-by-One
16 lines

Numbers of signal lines


between TCON(s) and panel

m-LVDS
112 lines

AiPi
64 lines

3.

Motion Picture Quality of 240Hz LCD

As noted, the 240Hz LCD module was developed to improve


motion picture quality. MPRT is a common method used to
quantify the amount of motion blur in a display. MPRT of the
240Hz LCD was measured and compared with other devices. For
best picture quality, a new overdriving method was also adopted.
To make the comparison, MPRT of 60 and 120Hz LCDs was
measured at various scroll speeds, and a 29-inch flat CRT TV was
also used for the test. MPRT measurements were based on blurred
edge time (BET), and the results are shown in figure 7. As the
figure shows, MPRT of the 240Hz LCD is 4.7ms at a scroll speed
of 12ppf, which is even better than that of the CRT. This is a
significant result, showing that LCD motion picture quality has
finally surpassed CRT performance.
20

MPRT (msec)

18

16.6

60Hz

15.7

15.7

8.6

9.4

16
14
12

11.0

120Hz

10
8

6.9

CRT

6.5
5.0

6
4

5.9

240Hz

4.7

4.7

12 ppf
Scroll Speed (pixel/frame)

16 ppf

2
0

8 ppf

Figure 7. MPRT of 60Hz/120Hz/240Hz LCDs and CRT


Figure 6. 240Hz ME/MC driving architecture (step 2)
The next step (step 2) is to implement the 240Hz system using a
single FRC and a single 240Hz TCON, as depicted in figure 6. In
the step 2 structure, no LVDS repeater is needed because a full
screen of 240Hz ME/MC data is processed in a single FRC chip.
Driving of the 240Hz panel is performed by a single TCON which
uses a V-by-One interface at its input and AiPi at its output.
Table 1 compares the number of signal lines required by miniLVDS and AiPi interfaces, and between existing LVDS and V-byOne interfaces. V-by-One HS is a new interface technology
developed by THine Electronics Co. Like AiPi, V-by-One uses an
embedded clock to eliminate clock-data skew, enabling greatly
increased bandwidth compared to LVDS. Since the 240Hz FRC

426 SID 09 DIGEST

4.

3D Display with 240Hz LCD

Additionally, in the proposed 240Hz LCD-TV, it is also possible


to deliver 3D functionality without any loss of resolution by using
LC shutter glasses and providing the left eye and right eye images
by time sequential driving as shown in figure 8. Conventionally,
120Hz LCD driving is not suitable for a 3D display using LC
shutter glasses because of crosstalk due to progressive scanning.
With the 240Hz LCD-TV, however, it is possible to minimize the
crosstalk between the left eye and right eye images by image
processing using twice as many frames as the 120Hz driving
scheme. Therefore, the 240Hz LCD-TV has many attractive
points for the consumer, as it not only provides superior motion
picture image quality, but it can also provide an uncompromised
3D experience using LC shutter glasses.

31.1 / S. S. Kim
Scanning
Left-eye data

Scanning
Black data

R
R

Scan
L
L

Scanning
Black data

Display
Left-eye data

Scan

Display
Right-eye data
Scan

Scan

Scanning
Right-eye data

Shutter glasses

Shutter glasses

(b) Display of left eye image

(a) Display of right eye image

Figure 8. 3D display based on 240Hz driving technology with LC shutter glasses

Figure 9. Worlds first 240Hz LCD-TV exhibited at Korea Electronics Show 2008
Table 2. Specifications of 240Hz LCD-TV
Resolution

1920 x 1080 (full HD)

Brightness

500cd/m2

Contrast Ratio

3,000:1 (static)

Color Gamut

72% of NTSC (CIE 1976)

Number of Colors

1.07B

Response Time

4.7ms (MPRT)

Viewing Angle

180 in all directions

5.

Conclusion

A true 240Hz FHD LCD-TV has been developed by using hG-2D


driving, the CS S-PVA pixel structure, and a 240Hz ME/MC
driving scheme based on a new interface. Pixel data are actually
updated 240 times per second, thereby avoiding ghost images and
other issues associated with so-called 240Hz hybrid techniques.
MPRT of the new 240Hz LCD-TV has been measured at 4.7ms,
providing a level of motion picture quality comparable to that of
CRTs. Moreover, this 240Hz LCD driving technology makes it
possible to deliver a full resolution 3D display by using LC shutter

glasses, thereby enabling 3D to become mainstream technology


for television. This 240Hz LCD-TV panel, the worlds first, was
demonstrated at IFA08 and at KES08 as shown in figure 9.
Specifications of the 240Hz panel are summarized in Table 2.

6.

References

[1] S. S. Kim, B. H. You, N. D. Kim, and B. H. Berkeley, Novel

[2]
[3]
[4]

[5]

120-Hz TFT-LCD motion-blur-reduction technology with


integrated motion-compensated frame-interpolation timing
controller, Journal of the SID 16/3, p. 403-413 (2008).
S. S. Kim, B. H. Berkeley, and T. S. Kim, Advancements
for highest performance LCD-TV, SID Symposium Digest
Tech. Papers 37, p. 1938-1941 (2006).
S. S. Kim, The worlds largest (82-in.) TFT-LCD, SID
Symposium Digest Tech. Papers 36, p. 1842-1847 (2005).
S. S. Kim, B. H. You, J. H. Cho, D. G. Kim, B. H. Berkeley,
and N. D. Kim, An 82-in. ultra-definition 120-Hz LCD TV
using new driving scheme and advanced Super PVA
technology, Journal of the SID 17/2, p. 71-78 (2009).
H. S. Nam et al., Cost effective 60Hz FHD LCD with
800Mbps AiPi Technology, SID Symposium Digest Tech.
Papers 39, p. 677-680 (2008).

SID 09 DIGEST 427

31.2 / D. Suzuki

31.2: Crosstalk-Free 3D Display with Time-Sequential OCB LCD


Daiichi Suzuki, Tetsuo Fukami, Emi Higano, Naoya Kubota, Toshiyuki Higano,
Seiji Kawaguchi, Yuuki Nishimoto, Kazuhiro Nishiyama, Kenji Nakao,
Research & Development Center, Toshiba Matsushita Display Technology Co., Ltd.,
26-2 Kawakita Nomi, Ishikawa, 923-1296, Japan

Takayoshi Tsukamoto, Hirofumi Kato


PC&AV-USE LCD Division, Toshiba Matsushita Display Technology Co., Ltd.,
1-9-2 Hatara Fukaya, Saitama, 366-0032, Japan

Abstract
We have developed high quality 3D display system with timesequential OCB LCD. That is 3D display of 15.4 inch WXGA
panel without the problem of 3D crosstalk, pseudoscopy and 3D
moir. Fast response OCB LCDs were used for both the panel
and active shutter glasses and improved blinking backlight
control technology was applied.

1.

This 3D system is full resolution display without any moir


because there is no barrier in the panel. Essentially wide viewing
angle properties are expected using active shutter glasses
regardless of viewing direction.
L

Image Source

WXGA
WXGA

Introduction

WXGA

WXGA

Recently the market of 3-dimentional (3D) system (stereoscopic


and auto-stereoscopic displays) has been rapidly expanding
worldwide. This is thought to be due to the increasing 3D
contents, such as 3D cinema. This has brought the demand for
3D displays. The various types of 3D displays have been
proposed. However, many 3D displays have big problems: loss of
spatial resolution, 3D crosstalk (ghosting), pseudoscopy (reverse
image) and 3D moir [1-5]. The types of 3D system using LCD
are mainly grouped into 2 types: one is area-division type and the
other is time-sequential. The former has the problem of the loss
of spatial resolution, occurrence of pseudoscopy, difficulty in
switching 2D/3D display, and 3D moir (especially in the type of
autostereoscopic display without 3D glasses). On the other
hand, the latter with active glasses has neither the loss of spatial
resolution, nor the 3D moir between pixels and barrier.
Therefore the time sequential system is suitable for high quality
3D display. However, this system can not be realized with
conventional LCDs, because their response is not fast enough.
Since the time-sequential 3D system requires higher frame rate
than 2D display, in the case of slow response, the luminance is
low and 3D crosstalk becomes worse.
Recently, we have developed OCB (Optically Compensated
Bend [6]) LCD with very fast response and realized high
performance of moving picture quality, MPRT=2.0ms, using
blinking backlight system[7-11].
In this paper we have applied the OCB LCD to the timesequential 3D display to realize high quality 3D display.

2.

Application of OCB to 3D system

2.1.

3D system

The 3D system we have developed is binocular vision system


and time-sequential system with active shutter glasses. In the
time-sequential system the images for each eye, left and right, are
alternately displayed in the LCD(Fig.1).The active shutter glasses
transmit the images synchronized with the display and 3D
image is created in our brain.

428 SID 09 DIGEST

WXGA

Image
for left eye

WXGA

Image
for right eye

LCD Panel
R

time

L
L
L

WXGA
WXGA

WXGA
WXGA

WXGA
WXGA
WXGA
WXGA

time

Active-shutter glasses

R
R
creation of one 3D image
from two 2D image

Persons Eye to Brain

Figure 1. Time-sequential 3D system with active shutter glasses.

2.2.

OCB technology for 3D system

The scheme and the principle of OCB time-sequential 3D


display system are shown in Fig.2. The 3D system consists of
OCB LCD display, OCB active shutter glasses and newly
developed blinking backlight control system.
3D system of time-sequential system requires for fast response
of LC because the frame frequency is twice as high as that of 2D
system. For example, the frame rate is 120Hz so the period of
one frame is only 8.33ms. The response time of LC should be
much shorter than 8.33ms. The slower response of LC shall
cause the 3D crosstalk and the loss of luminance.
In time-sequential 3D display, sequential images mixture causes
3D crosstalk. To reduce 3D crosstalk, we applied black insertion
driving and blinking backlight control technology. When the
display images are changed from right to left (or left to right), the
images are resetted to black state completly and the backlight is
turned off . Then the response of LC starts and reaches up to the
image data state. After that the backlight is turned on again. At
that time the previous images are not observed and only

ISSN/009-0966X/09/3901-0428-$1.00 2009 SID

31.2 / D. Suzuki
the correct images can be seen. In this way the reduction of the
images mixture can be realized and we can achieve small 3D
crosstalk.

OCB-panel
Transmittance

Backlight
ON

ON

ON

ON

The OCB mode is well-known for its fast response. We have


newly developed LC material to improve the response time of
OCB (Fig.3), r+d=3ms. The rise time from all gray level to
black is very fast, < 0.26ms (Fig.4). This means display image
can be resetted immediately and it make no influence to next
image. In this case very small 3D crosstalk is achieved. The fast
decay time from black to any gray level means high luminance of
3D display because of the high effective aperture in time range of
LC response. We can get high luminance 3D display using
OCB.
In case of other LC mode, large 3D-crosstalk will appear
because the response time to the black is not so fast. Also, the
brightness is very low because the time to the white is not fast.
The demand for 3D active shutter glasses is the same as that for
the LCD panel, that is, fast response of LC. The slow LC response
shall cause the 3D-crosstalk and shortage of luminance. So we
also applied OCB to active-shutter glasses.

Re sponse Time(ms)

OCB active shutter glasses


5

for Right Eye

4
3

OPEN CLOSE

OPEN

CLOSE

for Left Eye

CLOSE OPEN

CLOSE

OPEN

8 7

5 4

St art level

Black

Time

Whit e

End level

at room temperature

Figure 4. Response of OCB (gray to gray).

Figure 2. Scheme of OCB 3D-system.

1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

8
7
6
5
4
3
2
1

Whit e

Normalized Brightness

Worst
0.26ms

White Black

3.

Results of 3D qualitys evaluation

3.1.

Definition of 3D crosstalk

There are many quantitative parameters for 3D quality, such as


3D crosstalk, pseudoscopy and 3D moir. Among them the 3D
crosstalk is one of the most important parameter to discuss the 3D
quality. The 3D crosstalk means a ratio of incorrect, for opposite
eyes, images luminance to white luminance. Large 3D crosstalk
means that much incorrect image is visible and results in double
image (ghosting). That is obstacle to viewing 3D images and it
might introduce disturbing to visual fatigue.
0

1 2

3 4 5 6 7 8

9 10 11 12 13 14 15 16 17 18

Time (ms)
Figure 3.

Response of newly developed OCB


(r+d=3ms).

We have evaluated 3D quality in 3D crosstalk, pseudoscopy and


3D moir. We defined the 3D crosstalk ratio as below (shown in
Fig. 5 in detail).
3D crosstalk ratio (%) = (B-C)/(A) 100
A: the luminance through active shutter glass for left (or right) eye
when white level for both eyes is displayed in LCD panel.

SID 09 DIGEST 429

31.2 / D. Suzuki
B: the luminance through active shutter glass for left (or right) eye
when white level is displayed for opposite right (or left) eye,
black level is displayed for left (or right) eye in LCD panel.
C: the luminance through active shutter glass for left (or right) eye
when black level for both eyes is displayed in LCD panel.
According to the definition, pseudoscopy (reverse image) occurs
in the case over 50% of 3D crosstalk ratio.

panel

Of course the 3D moir does not occur at all. That is clearly


natural because there is no barrier (special lattice pattern) to create
3D pictures, which causes 3D moir. It is different from other
3D display (especially in auto-stereoscopic 3D system).

3.3.

Prototype specification

Table 1 shows the specification of the prototype 15.4 inch OCB


3D with time-sequential system. The definition of 3D-luminance
is the luminance through the active shutter glass when white level
is displayed in the LCD panel. The luminance of 3D image is
lower than the luminance of 2D image. However both of
brightness sense is about same, because the human eye adapts to
its viewing environment. The brightness sense becomes the
brightness in the adaptation field.

active-shutter
glasses
Luminance
meter

images displayed in the panel


Image for
Left eye

Image for
Right eye

White/White
(Luminance=A)

Black/White

3D crosstalk ratio(%

100
10
1

90
LCD

Black/Black
3D crosstalk ratio (%) = (B-C)/A 100

azimuth 45
azimuth 90
azimuth - 45

0.14%

0.1
0.08%
0.01
- 60 - 50 - 40 - 30 - 20 - 10 0 10 20 30 40 50 60
Polar Angledeg.

(Luminance=B)

(Luminance=C)

azimuth 0 - 180

Figure 6. Measurement result of 3D crosstalk ratio.


Table 1. Specification of 15.4 inch OCB 3D system

Specification
Figure 5.

3.2.

The definition of 3D crosstalk ratio.

Experimental Results

We have measured 3D crosstalk ratio in all viewing angle. We


used the luminance meter and goniometer to measure the
characteristics of viewing angle. The active shutter glass is set
in the front of the luminance meter. The experimental result is in
Fig. 6. This is the data of 15.4 inch WXGA OCB panel. The 3D
crosstalk ratio is only 0.08% of the normal direction at the center
of the panel. This value is very low and almost the same as the
visibility thresholds [12].
Angular dependence of the 3D crosstalk ratio is almost the same
and less than 0.14% in wide viewing angle. At the same time,
there is no pseudoscopy.
We would like to declare that our panel is the "crosstalk-free"
3D display system.

430 SID 09 DIGEST

Diagonal Size

15.4inch

Pixel Format

1280RGB800

2D-Luminance

400 cd/m2

3D-Luminance

120 cd/m2

3D-Crosstalk ratio

0.08% at normal direction


0.14% in polar 60

Pseudoscopy

no occurrence
from all viewing angle

3D Moir

no occurrence
from all viewing angle

the luminance through active shutter glass

31.2 / D. Suzuki
4.

Conclusion

Horikoshi, H. Ujike, Measurement of Multi-view and


Integral Photography Displays Based on Sampling in Ray
Space,IDW'08 Proc., pp.1115-1118, 2008.

We have realized high quality 3D display system using timesequential OCB LCD. This is 3D crosstalk-free, no pseudoscopy,
no 3D moir in wide viewing angle. It was achieved by fast
response OCB panel, fast response OCB active shutter glasses and
newly developed blinking backlight control.

[6] T. Miyashita, P. Vetter, M. Suzuki, Y. Yamaguchi, T.

5.

[7] K. Kumagawa, A. Takimoto, H.Wakemoto, Fast Response

References

[1] G. Hamagishi, K. Taira, K. Izumi, S. Uehara, T. Nomura, K.


Mashitani, A. Miyazawa, T. Koike, A. Yuuki, T. Horikoshi,
Y. Yoshihara, Y. Hisatake, H. Ujike, Y. Nakano,
Ergonomics for 3D Displays and Their Standardization,
IDW'08 Proc., pp.1099-1102, 2008.

[2] K. Taira, G. Hamagishi, K. Izumi, S. Uehara, T. Nomura, K.


Mashitani, A. Miyazawa, T. Koike, A. Yuuki, T. Horikoshi,
Y. Hisatake, H. Ujike, Y. Nakano, Variation of
Autostereoscopic Displays and Their Measurement,IDW'08
Proc., pp.1103-1106, 2008.

[3] S. Uehara, K. Taira, G. Hamagishi, K. Izumi, T. Nomura, K.


Mashitani, A. Miyazawa, T. Koike, A. Yuuki, T. Horikoshi,
Y. Hisatake, H. Ujike, Methodology of Optical Mesurement
for Autostereoscopic Displays, IDW'08 Proc., pp.11071110, 2008.

[4] A. Yuuki, S. Uehara, K. Taira, G. Hamagishi, K. Izumi, T.


Nomura, K. Mashitani, A. Miyazawa, T. Koike, T.
Horikoshi, H. Ujike, Viewing Zones of Autostereoscopic
Displays and their Measurement Methods, IDW'08 Proc.,
pp.1111-1114, 2008.

[5] T. Koike, A. Yuuki, S. Uehara, K. Taira, G. Hamagishi, K.


Izumi, T. Nomura, K. Mashitani, A. Miyazawa, T.

Uchida, Wide Viewing Angle Display Mode for Active


Matrix LCD Using Bend Alignment Liquid Crystal Cell,
Eurodisplay93 Conf. Proc., pp.149-152, 1993.
OCB-LCD for TV Applications SID02 Symposium Digest,
pp.1288-1291, 2002.

[8] K. Nakao, D. Suzuki, T. Kojima, M. Tsukane, H. Wakemoto,


High-Speed Bend Transition Method using Electrical Twist
Field in OCB Mode TFT-LCDs, SID04 Symposium
Digest, pp.1416-1419, 2004.

[9] K. Nishiyama, M. Okita, S. Kawaguchi, K. Teranishi, R.


Takamatsu, 32WXGA LCD TV using OCB Mode, Low
Temperature p-Si TFT and Blinking Backlight Technology
SID05 Symposium Digest, pp.132-135, 2005.

[10] K. Nakao, Y. Tanaka, H. Takahara, S. Kawaguchi, K.


Nishiyama, S. Araki, A. Takimoto, OCB-LCDs with New
Driving Method Having Fast Response of 2.3-ms MPRT and
High Contrast Ratio of 1000:1, SID07 Symposium Digest,
pp.138-141, 2007.

[11] S. Araki, K. Nakao, S. Kawaguchi, Y. Nishimoto,


K.Nishiyama, K. Shiiba, A.Takimoto, High Contrast, Low
MPRT OCB LCD with Dynamic Backlight Control
Technology, SID08 Symposium Digest, pp.975-978, 2008

[12] S.Pastoor, Human Factors of 3D Imaging: Results of Recent


Research at Helnrich-Hertz-Institut Berlin, IDW95 Proc.,
pp.69-72, 1995

SID 09 DIGEST 431

31.3 / S. Uehara

31.3: Reduction and Measurement of 3D Moir


Caused by Lenticular Sheet and Backlight
Shin-ichi Uehara, Tsutomu Hiroya, Kouji Shigemura, and Hideki Asada
NEC LCD Technologies, Ltd., Kawasaki, Kanagawa, Japan

Abstract
We found that interference between a 3D display's lenticular sheet
and backlight prism sheet causes 3D moir. This previously
unknown phenomenon affects image quality. Its mechanism,
reduction methods and their implementation results are
described. An appropriate measurement method for 3D moir is
also proposed.

1.

Introduction

With the improvement in the performance of displays, such as


LCDs and PDPs, a variety of 3D displays not requiring special
glasses, that are often known as autostereoscopic displays, have
recently been developed for entertainment, medical, design, and
other applications. In order to show the performance of
autostereoscopic displays, the certification of space for viewing
3D images is considered to be essential. This space is called as
QSVS (Qualified Stereoscopic Viewing Space) [1] [2], and 3D
moir is one of the significant factors that determine the QSVS. In
order to evaluate the 3D moir correctly, careful consideration of
the cause of 3D moir is required, otherwise the performance
would not be shown correctly, thus misleading the development
of 3D displays, such as by improvement of undesirable
performance or meaningless competition in specifications.
In general, 3D moir is considered to be caused by the
combination of the optical components, e.g. a lenticular, and a
pixel structure, e.g. black matrix. In addition, in our previous
work, we found that a combination of a reflector with a rough
surface and a lenticular causes 3D moir in reflective displays [3].
In this paper, we describe a previously unknown 3D moir caused
by interference between the backlight structure and a lenticular
sheet for 3D display. We also describe its reduction methods and
their implementation result. We propose appropriate measurement
methods for 3D moir, which are especially important for
establishment of the international standards, which are being
developed by many organizations for standardization, such as
ISO, IEC, ICDM and so on.

2.

3D moir and its reduction methods

2.1.

3D moir

3D moir is generated by an interference of the optics for


autostereoscopy, such as a lenticular, and a display structure. 3D
moir is considered to be a kind of moir in a wide sense. When
its spatial frequency is high, e.g. in the case of slant lenticular, it
looks like a moir on ordinary 2D displays, which often appear
when layers of grids are overlapped. The moir on 2D displays is
patterns of intensity variations superimposed on the screen image.
For example, ISO9241-303 defines that moir is a regular image
superimposed on the intended image, and that they can appear as
ripples, waves and intensity variations that are superimposed on
the screen image. In this case, conventional measurement methods
can be applied.
However, in some cases, the spatial frequency of a 3D moir
will be lower, and its appearance is quite different from that of
the moir on ordinary 2D displays. In this case, luminance
angular fluctuation increases, causing the uniformity on the
screen to be degraded, that is, what is called a black band or
banding.
3D moir is considered to be a complex phenomenon of various
spatial frequencies induced by various factors of 3D displays.
Since autostereoscopic displays have the function of parallax
view, the optical components and the display structures make
3D moir unavoidable. In order to consider the factors of 3D
moir, the complex refractive index is effective, as shown in
Figure 1.
(a) The imaginary part indicates the absorption, which is
related to the black matrix. Interference between the black
matrix and the optical components sometimes causes
luminance 3D moir, because the black matrix doesnt emit
light.

n( ) + ik ( )

Lenticular Lens

Reflector

Color
Filters
Black Matrix
(a) Absorption
by Black Matrix

432 SID 09 DIGEST

(b) Absorption Dispersion


by Color Filters

(c) Reflection by Reflector

Figure 1. Factors of 3D moir

ISSN/009-0966X/09/3901-0432-$1.00 2009 SID

Prism Sheet (Backlight)


(d) Refraction
on Prism Sheet

31.3 / S. Uehara

Lenticular
Right-eye
Pixel

Dark

Left-eye
Pixel

LCD panel
Left Slope

Bright

Right Slope

Figure 3.
Optical Distributions
of Slopes on Prism

Prism Sheet (Backlight)


Figure 2. Structure of 3D LCD
with Lenticular and Backlight

Figure 5. Photo of 3D moir Caused by Lenticular and


Backlight
(b) When the imaginary part has the dispersion in wavelength,
such as color filters, chromaticity variations, called
chromaticity 3D moir, sometimes occur. In addition, when
the array direction of same colors is not parallel to the
direction of lens effect, chromaticity 3D moir will be
undesirable.
(c) The real part represents the refraction and the reflection,
including the direction of rays. Reflection is related to the
rough reflector. In our previous work, we found that a
combination of the micro structure on the reflector and a
lenticular lens gives rise to luminance 3D moir. In order to
avoid this, we optimized the micro structure and the
lenticular lens, and then we have developed a high-quality
reflective type [3] and a transflective type of 2D and 3D
display [4].
(d) Refraction is related to backlight structure, which is the main
topic described in this paper. In addition, the 3D moir
caused by a lenticular and a backlight is also interesting in
that its factors are situated on both of the outer sides of the
LCD panel.
The structure of a 3D LCD with a lenticular and a backlight is
shown in Figure 2. In general, a backlight has various optical
sheets, such as light guide or prism sheets, which have roughness
patterns on the surface to idealize the optical property. These
patterns are composed of many slopes of different angles whose
light distributions are different. This means that the optical sheet
is a gathering of various different distributions microscopically.

(a)

(b)

Figure 4. Mechanism of 3D moir


Caused by Lenticular and Backlight

BL with lens
- 3D moir appears

BL without lens
- no 3D moir

Figure 6. 3D moir by BL with Lens


For example, as shown in Figure 3, the prism sheet has two kinds
of slopes whose angles of inclination are different, which means
that the optical distribution on the left slope of each prism is
different from that on the right slope. In addition, the light output
from each slope is non-lambertian.
Therefore, a lens-type 3D display has a close relation between the
angle of light output from a lens and a region on the prism, such
as the left or right slopes. As shown in Figure 4(a) and Figure
4(b), when the region which corresponds to an angle of parallel
light is narrower than the prism pitch, since the distribution will
be changed according to the position of the region, the angular
luminance will fluctuate, thus causing 3D moir. Figure 5 shows a
photo of this 3D moir with an LCD panel.
In Figure 6, although the LCD panel was removed, the 3D moir
didnt disappear, that means the 3D moir is caused by a
lenticular and a backlight. The 3D moir will be more noticeable,
when the LCD module becomes thinner, that is to say, when the
distance between the prism sheet and the focal plane of a lens
becomes shorter. The pitch of a lens or prism is also an impact
parameter. Recently, because not only small or middle size
displays, but also large size displays become thinner, and because
backlights that use many optical sheets are becoming popular, this
3D moir will be a great issue in the near future.

SID 09 DIGEST 433

31.3 / S. Uehara

Diffusive
Layer

V<LxS/f
V
(c) Introduction of
(a) Optimization of
(b) Inclination of
Diffusive Layer
Prism Pitch
Prism Sheet
Figure 7. Reduction Methods for 3D
moir

Figure 8. Implementation Result of


Our Methods

Lenticular
Lens

Square
Red
Green
Blue
Right-Eye Pixel

Left-Eye Pixel

Figure 9. HDDP Arrangement

2.2.

Figure 10. Photo of Sample Image

Reduction methods for 3D moir

In order to reduce the 3D moir, we propose three methods, as


shown in Figure 7. The first method is to optimize the prism pitch
V, the lens pitch L, the focal length f and the distance S between
LCD panel and backlight as follows:
V<LxS/f

(1)

The second method is to incline the prism sheet against the


lenticular. The third method is to introduce a diffusive layer
between LCD panel and backlight. These three methods had
better be used together, because the effect of each method is
limited. Figure 8 shows a photo of our implementation result,
which shows these methods are effective.

3.

Implementation and measurement

3.1.

Application to HDDP arrangement

We have also applied the reduction methods to our 2D/3D-LCD


with novel pixel structure, called HDDP (Horizontally DoubleDensity Pixels) [5], as shown in Figure 9. In the HDDP
arrangement, since it incorporates rectangular pixels whose width
is half that of their height, in a lenticular-lens equipped 3D mode,
the horizontal resolution will equal that of the vertical. This not
only results in high 3D image quality, it also means that 2D
characters can be displayed with perfect legibility.

434 SID 09 DIGEST

We have developed 2.7-inch diagonal 2D/3D LCD with the


methods. Each pixel set consists of 2 pixels (for left and right),
and each pixel consists of 3 RGB dots. Horizontal and vertical
pixel set numbers are, respectively, 240 and 320 (QVGA format).
Thanks to the fully custom-made LCD for autostereoscopic
display and the optimized optics, the 3D crosstalk level is 0.8%,
which is truly excellent for viewing. The maximum width of the
QSVS is about 50mm at the viewing distance of 380mm. Figure
10 shows a photo of a sample image for the 2D/3D LCD, which
shows high image quality without patterns of intensity variations.

3.2.

Measurement methods for 3D moire and


their implementation results

As mentioned above, since 3D moir occurs in various spatial


frequencies according to many factors, the measurement methods
based on the principle of 3D moir should be established. This
will enable the method to correspond to an unexpected 3D moir.
We consider that not only the spatial measurement using a
calibrated camera, but also the angular measurement is significant
to certify the viewing space. The ordinary moir can be measured
by conventional methods. In this paper, the angular measurement
is described.

31.3 / S. Uehara

Luminance

LA

250

LB

Angle

Threshold Curve

LA LB

Angular
Difference

( LA + LB )

A B

Luminance[cd/m2]

Cm =

-10

-8

-6

-4

-2
0
2
Angle [degree]

10

300
200
100
0
-20

-10
0
10
20
Angle[degree]
Figure 12. Measurement Result of 3D Luminance

Figure 11. 3D moir


a) Objective
The objective of this measurement is to obtain the viewing space
where the requirement for 3D moir is fulfilled. In particular, the
low-frequency 3D moir peculiar to autostereoscopic displays that
causes luminance non-uniformity or luminance difference is
measured by optical measurement.
b) Set-up and procedure
Luminance profile in all views white is measured at the center of
the screen, otherwise luminance profiles of each view are
measured, and all luminance profiles are summed up as all white
luminance profile.
c) Analysis
[1] On luminance profile in all white, inflection points, which are
points where the curvature changes sign, are detected, as shown in
Figure 11.
[2] Then, luminance contrast modulation and angular differences
are calculated between two neighboring inflection points. The
luminance contrast modulation is
Cm = | LA LB | / ( LA + LB )

Angular
Difference: 3.2

200

B A

Luminance Contrast
Modulation

Luminance Contrast
Modulation: 5.5%

300

Luminance [cd/m2]
. .

(2)

The luminance contrast modulation is also called Michelson


contrast, that shows the difference from the average.
[3] These values should be below the curve of threshold
characteristics for 3D moir. This threshold curve should be
determined carefully based on ergonomics data. Furthermore, it
should be confirmed whether the threshold depends on the
luminance level, viewing distance, and a number of inflection
points in each lobe. As a simple method, the value of Cm / | A- B|
will be effective.
The minimal and essential location is considered to be the
center of the screen, because the center is mainly used. In
addition, the left and the right of the screen had better be used for
measurement, because some autostereoscopic displays have
different optical conditions in each location.
As regards chromaticity 3D moir, luminance measurement will
be replaced by chromaticity measurement.

Figure 12 shows the measurement results of our 2D/3D LCD.


Thanks to the reduction methods, the luminance profile is very
smooth. The maximum luminance contrast modulation is 5.5% at
the angular difference of 3.2 degrees. Although the threshold
should be determined carefully based on ergonomics, the
subjective evaluation shows this level has no problem for
viewing. Figure 10 also shows that the ordinary moir doesnt
occur.

4.

Summary

We found interference between a 3D display's lenticular sheet and


backlight optics causes 3D moir, which is a previously unknown
phenomenon that affects image quality. The thinner an LCD
module is, the more noticeable and unavoidable this 3D moir is.
Its mechanism, reduction methods and their implementation
results were described. We also proposed the measurement
methods based on the principle of 3D moir.

5.

Acknowledgements

The authors thank JENC (Japanese Ergonomics National


Committee) / TC 159 / SC 4 / WG 2 as Japanese mirror
committee for ISO, and also thank ELDIM for providing the
measurement data and fruitful discussions.

6.

References

[1] K. Taira, G. Hamagishi, K. Izumi, S. Uehara, T. Nomura, K.

[2]

[3]
[4]
[5]

Mashitani, A. Miyazawa, T. Koike, A. Yuuki, T. Horikoshi,


Y. Hisatake, H. Ujike, and Y. Nakano, Proc. of the 15th
IDW, 3D2-2, pp.1103-1106 (2008).
S. Uehara, K. Taira, G. Hamagishi, A.Yuuki, K. Mashitani,
T. Koike, K. Izumi, T. Nomura, A. Miyazawa, T. Horikoshi,
and H. Ujike, Proc. of the 15th IDW, 3D2-3, pp.1107-1110
(2008).
S. Uehara, et. al., Proc. of 26th IDRC, pp.131-134 (2006).
S. Uehara, et. al., 6803-22, Proc. of SPIE-IS&T Electronic
Imaging, SPIE Vol.6803 (2008).
S. Uehara, et. al., Asia Display / IMID'04 Digest, pp.783-786
(2004).

SID 09 DIGEST 435

31.4 / H.-H. Huang

31.4: Direct View and Projection Switchable Mobile Display


Hsin-Hsuan Huang and Cheng-Huan Chen
Dept. of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan, R.O.C

Yu-Cheng Lai , Yu-Cheng Chang , Yi-Pai Huang and Han-Ping D. Shieh


Dept. of Photonics and Display Institute, National Chiao Tung University, Hsinchu, Taiwan, R.O.C.

Chih-Wen Chen
LC Technology Dept., AC Technology Division, AU Optronics Corporation, Hsinchu, Taiwan, R.O.C

Abstract
The function of mobile phone becomes more versatile as a personal
assistant. In order to make display device optimized for different
function of the mobile phone, this paper proposes a display device
which can be switched among 2D direct view, 3D direct view and
projection modes while using single liquid crystal(LC) panel as
image source. The device features a highly collimated backlight
unit and polymer dispersed liquid crystal(PDLC) diffuser at the
back of the LC panel, and a switchable lenticular array as well as
a stretchable projection lens in front. The ray tracing simulation
has shown well controlled light path of the backlight module and
sufficient image quality of the projection lens while maintaining
the compactness of the device for mobile application.

1.

Introduction

The function of mobile phone nowadays is not limited for speech


communication as for its original purpose and becomes more
versatile as a personal assistant and/or entertainment unit. The
integration of camera imaging, PDA, GPS and video game has
already been popular in the market. Integration of TV and internet
access etc will gain more attention in the near future. The display
unit plays the key role as the human interface for those different
function and application of the mobile phone. However, different
application requests different specification of the display unit with

the consideration on effectiveness and efficiency. Therefore, quite


a few integrated or switchable display unit have been proposed,
including 2D/3D switchable display[1] and pico projector
embedded into a handset with a direct view panel of its own[2]
etc. In this paper, a versatile display unit which can be switched
among 2D direct view, 3D direct view and projection modes while
using single LC panel as the image source has been proposed. The
form factor is also taken into consideration in evaluating the
feasibility of the architecture in order to ensure its availability as a
mobile display unit.

2.

Switchable display configuration

The architecture of the display unit being proposed is shown in


Figure 1, which looks similar to a folding mobile phone with the
image source panel and illumination backlight module in the
bottom plate, and a projection lens in the top plate. The image
source is a traditional liquid crystal panel with a polymer
dispersed liquid crystal(PDLC) diffuser[3] underneath in between
the backlight module and the LC panel. The backlight module
needs to have the emerging light from the top surface as
collimated as possible. A switchable lenticular array[4] or parallax
barrier[5] is attached onto the top surface of the LC panel. The
projection lens in the top plate is a stretchable one, similar to
several camera lenses, which stretch out only when being used.

Figure 1 Configuration of the proposed three-in-one hybrid display unit

436 SID 09 DIGEST

ISSN/009-0966X/09/3901-0436-$1.00 2009 SID

31.4 / H.-H. Huang


Three different operation modes of the display unit, namely 2D direct
view, 3D direct view and projection modes, are illustrated in Figure 2.
For 2D direct view as in Figure 2(a), the cover plate is flipped over, the
PDLC is off and becomes highly diffusive, and the switchable lenticular
array is off to behave like a simple transparent plate. In 3D direct view
mode as shown in Figure 2(b), the cover plate is flipped over, the PDLC
is on to become a simple transparent film, the switchable lenticular array
is on so as to direct light towards different viewing direction for different
group of pixel and become a spatial multiplexed autostereoscopic 3D
display. The projection mode requests to close the cover plate and
stretch the projection lens element, as shown in Figure 2(c). In this
mode, the PDLC is on to make the illumination highly collimated, and
the switchable lenticular array should be disabled.

distribution over the light guide surface and the angular profile of
emerging light respectively from the simulation result of ray tracing
using Light Tools. The uniformity reaches 70% and the angular
divergence is less than 5 measured at half intensity. The total
thickness of the backlight light module is 15mm.

(a)

(b)
(a)

(b)

Figure 3 Configuration of high collimation backlight


module(a) Side view (b)Birds eye view

(c)
Figure.2 Operation mode of hybrid display unit
(a) 2D (b) 3D (c) Projection mode

3.

(a)

High collimation backlight module

The above mentioned concept is clearly feasible based on the


currently available technology, and the key issues become the form
factor and maintaining efficiency in switching between direct view
and projection mode. A highly collimated backlight is desired for
projection mode, while it can be diffused in direct view mode by a
PDLC diffuser at off state. Figure 3 shows the top view and birds
eye view of the high collimation backlight module using LEDs as
the light source. The LEDs located at the side edge of the light guide
are firstly collimated by a parabolic reflector, one for each LED.
Micro-scale prism structure is patterned on the bottom surface of the
light guide to make the emerging light from the top surface confined
within a small cone angle. Figure 4(a) and (b) show the illuminance

(b)
Figure 4 (a)Illuminace distribution (b)Angular profile of
emerging light from backlight module Stretchable projection
lens

SID 09 DIGEST 437

31.4 / H.-H. Huang


4.

Stretchable projection lens

The projection lens needs to provide good quality image while


having compact form factor. The trade-off can be resolved with a
stretchable lens module as what has been adapted in several
compact digital cameras. Figure 5(a) show the layout and light
path of the projection lens at stretching condition. All the four
lenses are spherical plastic lens. Table 1 lists the parameters of
projection lens which include radius, thickness and the lens
material. PMMA is the only material used in this design because
of cost consideration. The panel size on the right end of the light
path has a diagonal size of 1.96 inches and an aspect ratio of 4:3.
The pixel size is 0.1245mm, which corresponds to spatial
frequency of 4.0 lp/mm. Figure 5(b) shows the MTF of the
projection lens, which indicates a MTF of 50% at 5 lp/mm. One
major feature of the projection lens is telecentricity at the panel
side with a ray cone of 5 half angle. Accompanying with the
high collimation light guide, it makes the best use of the energy
from the light source in projection mode. The other feature of the
projection lens is the compact size of lens element stretching far
away from the image source, which makes the display unit
mechanically stable in projection mode operation. The total track
of the light path in Figure 5(a) is 500mm, and the total thickness
of the lens elements is 19.9mm, which will be roughly the
thickness of the cover plate containing projection lens elements.
The project image size is 180mm*135mm, roughly A4 size, with a
throw distance of 427mm.

(a)

Table 1 the list of radius, thickness and material for the


projection lens

5.

Radius

Thickness

Object

Infinity

426.7

16.5

3.6

-465.4

Stop

Infinity

3.5

-16.2

1.1

19

2.6

41.3

3.8

-23.7

29.6

50.5

11.4

-656.9

12.6

Image

Infinity

Glass
PMMA
(b)
PMMA
PMMA

6.
PMMA

Conclusion

A three-in-one hybrid display unit using single image source panel


has been proposed as the visual interface for mobile device. The
display provides switchable function among 2D direct view, 3D
direct view and projection. The major feature of the hardware is a
highly collimated backlight module and a stretchable projection
lens. Ray tracing simulation has shown sufficient performance of
both modules and the feasibility of the proposed display
architecture.

438 SID 09 DIGEST

Figure 5. (a) Light path and layout (b) MTF of projection lens
at stretching condition

References

[1] K. J. Kim, H. Kang, " Development of a 42-in. 2-D/3-D


switchable display using multi-view technology for publicinformation-display applications," in SID 15/11, 2007,
pp.899-903
[2] Denis Darmon, John R. McNeil, and Mark A. Handschy "
LED-Illuminated Pico Projector Architectures," J. Appl.
Phys. Lett. Vol.48, pp.269-271,1986.
[3] W. Doane, N.A. Vaz, B.-G. Wu, S. Zumer, " Field controlled
light scattering from nematic microdroplets," Prodeedings of
SID, 2008,pp.1070-1073.
[4] S.T. de Zwart, " A 20 Switchable Auto-Stereoscopic 2D/3D
Display," Prodeedings of IDW04,pp.1459-1460.
[5] Byoungho Lee, Heejin Choi, Joohwan Kim, Yunhee Kim,
and Seong-Woo Cho, " Status and Prospects of
Autostereoscopic 3D Display Technologies," Prodeedings of
IEEE, pp. 354-355, 2007.