The Development of Luminance Uniformity Measurement for CNT-BLU
Based on Human Visual Perception
Kuo-Hao Tang, Yueh-Hua Lee, FengChiaUniversity
Abstract--CNT-BLU (Carbon NanoTube backlight unit), an emerging backlight component for LCD, has several advantages such as light weight, superior color performance etc, and therefore have the potential to replace traditional CCFL (Cold Cathode Fluorescent Lamps)backlight unit. However, the current uniformity of CNT-BLU still can not compete with that of CCFL due to immaturemanufacturing process. The CNT-BLUs under current technology limits producedotted appearance, and thus prevent it from using existing luminance uniformity measurements such as VESA or ISO Standard, which is more suitable for measuring CCFL-LCD with relatively good uniformity. This study developed a new method, Line Uniformity,to monitor the uniformity improvement for CNT-BLU before it can be accepted by the market. This method was compared with VESA and U Formula with respect to human perception. A set of CNT-BLU images with different levels of dotted appearances was presented to 15 participants. The subjective acceptance threshold for images with the same level of dotted appearances was then calculated. The uniformities for VESA, U Formula, and Line Uniformity were calculated and fitted to the subjective acceptance threshold separately. The results showed that Line Uniformity was better fitted to acceptance threshold with R2 ranging from 0.80 to 0.92 whereas VESA and U Formula generatedR2 ranging only from 0.00 to 0.49.
Keywords--CNT-BLU,human perception, luminance uniformity, threshold.
I.INTRODUCTION
As the display size increases, the cost associated with LCD backlight units (BLU) becomes larger.When compared with CCFL backlight units, CNT(Carbon Nanotube)backlight unitshave the advantages of less power consumption, superior color performance, and no toxic chemicals [8].Under the RoHS (Restriction of Hazardous Substances) regulations, CCFL-BLU was restricted from 2006 July 1th due to the toxic mercury (Hg) gas. In addition, without the optical film and diffuser that used in CCFL-BLU, CNT-BLU can save up to 60% of the BLU cost. Hence, CNT-BLU has been considered asa potential candidate that may replace CCFL-BLU for the next generation of large screen LCD.
Carbon Nanotube (CNT) was discovered by Iijima in 1991 and since then many studies focus on the properties of CNT, particularly the field emission effect and its applications on lighting, back light unitfor LCD, or field emission display (FED) [5],[7]. The light emission principleof CNT-BLU are similar to that of CRT (Cathode Ray Tube) display, both based on tunneling of electrons through the surface potential barrier. In stead of using a single emitting source like CRT, CNT-BLUuses an array of emitters, and each emitter is corresponding to each pixel. Thus, the distance between cathode and anode can be effectively shortened, which makes the flat products possible.
In terms of manufacturing process, mass production of CNT-BLU is still under development and some issues are yet to be addressed, such as of luminance uniformity, light upcontrol, material and manufacturing costs etc.Luminance uniformity isa very important criterion for CNT-BLU.It relates to the end users’ perception and it also affects the grading and pricing strategy of the product. However, a variety of physical factors in the manufacturing process such as non-uniformly distributed CNT material and foreign particleswithin the CNT material may cause luminance uniformity related problems. When compared with mainstream flat panel display products such as LCDs, CNT-BLU shows different appearance due to suchmanufacturingflaws. Mura[1]in LCD is broadly discussed and has three different pattern types, spot Mura, line Mura and region Mura [3]. There are two categories of luminance uniformity phenomenon for CNT-BLU, the first one is mottled background defect, and the second one ispattern defect. From Fig. 1, the mottled background defectis a randomly dotted background with different shadings of spots, representing a typical BLU image under current manufacturing processes. On the other hand, pattern defectare very similar to LCD Mura, which also include lineand region in shape emerged from mottled background as shown in Fig. 2.
In order to measure luminance uniformity, four widely usedluminance uniformity measurements arepresented below(1)-(4). These methods are based on sampling multiple points and only consider the maximal and minimal luminance among sampled points.
VESA： / Non-Uniformity=[1-(Lmin/Lmax)]*100% / （1）ISO： / Uniformity=(Lmax/Lmin) / （2）
SPWG： / Uniformity=[(Lmax-Lmin) / Lmax] *100% / （3）
TCO： / Luminance Variation = (Lmax/Lmin). / （4）
Where Lmaxand Lmin represent the maximal and minimal luminance value among measured points.
Fig.1. CNT-BLU with pattern defect: line, region defects with a parti-colored background / Fig. 2.20’ CNT-BLU without patterned defect, and has a parti-colored background / Fig. 3.23’ CCFL BLUFig.4. Six measured lines in Line Uniformity method
Among these standards, VESA standard is widely recognized because it provides a comprehensive catalogof versatile optical measurements and informativetechnical discussions well-groundedin solid metrology[2]. A total of 9 or 16 points are required to be measuredaccording to the size of the panel when using VESA standard.Such sampling measurement of luminance is quite risky for CNT-BLU, which has obvious mottled appearance.The use of these measuring methods, consideringonlythe minimal and maximal luminance, may not reflect the true luminance uniformity.
In order to calculate the luminance uniformity more precisely, Reference [6] suggested that luminance uniformity measurement could be based on line measurement, which is, measuring luminance pixel by pixel on a line across the panel. They also proposed that the ratio of minimal and maximal luminance of measuring points should not be under 0.7 to accepta panel. The idea of cross section measurement was mentioned in another research. Reference [1] provided equation(5) that the average and standard deviation of luminance of measured lines were used to represent the LED-BLU uniformity.
/ （5）Where σL and Lave stands for standard deviation and average of measuring luminance respectively.
From Fig. 3, it can be seen that a CCFL-BLU has a much better luminance uniformitycompared with CNT-BLU shown in Fig. 2. The difference may explain why VESA being suitable for CCFL-BLU, may not be a good measure for CNT-BLU, especially during its research and development stage. The purpose of this study is to develop a more appropriatemethod to depict the luminance uniformity for CNT-BLU withmottled backgrounddefect (for pattern defect, please seereference [10]) and compare it with VESA Standard and U Formula, in order to evaluate the consistency with human perception.
II.Development of Line Uniformity
In order to present the luminance uniformity of the mottled background of a CNT-BLU, Line Uniformity was proposed. Six lines are to be measured for Line Uniformity. This idea comes from the 9 points measurement for LCD according to VESA standard. The locations of these 9 points are as shown in Fig.4. The 3 points on the left side are located at 1/6 of the BLU width from the left edge. The 3 top points are located at 1/6 of the BLU height from the top edge. The 3 points on the right side and 3 bottom points are symmetric with respect to the central vertical and horizontal lines. Six linesare connected through these 9 points as shown in Fig.4.
The luminance of each pixel on all the 6 lineare measured and the sum of the absolute values of the difference between the two adjacent pixels across all six lines is calculated to show the Line Uniformity as shown in (6). The larger the Line uniformity, the worse the luminance uniformity of the panel is. With this equation, the luminance variance between two adjacent pixels canbe detected.
/ (6)Where L(i,j) stands for the change of luminance value of jth pixel on ith line, and nistands for the number of pixel on ith line.
III.Development of CNT-BLU Image Model
The BLU is a module providing light source for an LCD panel, which allows less than 10% of light from BLU passing through. Therefore, the meaningful way to measure the luminance uniformity fora CNT-BLU is after it being assembled with an LCD panel. Due to that the CNT-BLU manufacturing is stillin pilot run stage, the numbers of available BLUwere small, and even less being assembled with an LCD panel. To cope with this limitation, a CNT-BLU image model was developed for this study to present simulated CNT-BLU images on a regular LCD with relatively high fidelity.
A.Transformation from CNT-BLU to LCD with CNT-BLU
The construction of transformation equation between the luminance of CNT-BLU and that of assembled with LCDpanel was conducted ina major research institute in Taiwan.The measuring environment was controlled at luminance less than 1 lux, temperature at 25Celsius degree,humidity at 35%, and gate voltage of CNT-BLU controlled to a fixed level at 340 volts.
Since the luminance of CNT-BLU changes according to the anode current, in order to cover a more complete luminance range, anode current was adjusted from 2mA to 22mA with 1mA per step, the corresponding measured luminance of CNT-BLU increased from 22.04 nits to 1223.05 nits. There were 650*490 points measured in a panel and 21 steps in terms of current control generating 6,688,500 (21*650*490) luminance measurementsfor building the transformation equation between the luminance of CNT-BLU and that of CNT-BLU with LCDpanel.Regression modal was used to fit the measured data and the fitted model (R2 = 0.98)was shown inFig.5 (left).
B.Transformation from LCD with CNT-BLUto a regular LCD
To reproduce the image from an LCD with CNT-BLU onto a regular LCD display, Acer AL917, a 19 inches LCD was first chosen as an experimental apparatus. Given a fixed brightness and contrast setting for this LCD, the functionbetween the luminance measured from this LCD and the gray levelof the panel, i.e.,from RGB(0, 0, 0) to RGB(255, 255, 255),can be determined(R2 = 0.99)as shown in Fig.5 (right).
Eighteen original CNT-BLUswere used in this study. These images of CNT-BLUs were taken by using ProMetric system where the luminance of each CNT-BLU pixel can be recorded. Then with these two functions, these images of CNT-BLU can be converted to predicted luminance as if an LCD panel was assembled, then finally transformed to a RGB value representing the corresponding luminance.
As we mentioned before, the current quality of CNT-BLU is not up to the market, and the images of the CNT-BLUs were far below users’ acceptance threshold. To simulate the future improvement of the image quality and still keep the characteristics of CNT-BLU, the Gaussian Blur function [9]was used to mimic this improvement process. For each of the 18 images, 11 blurring levels were generated. Level 1 was the original image and the Level 11 had a uniformity similar to a normal LCD.
C.Subjects and Experimental Task
Fifteen engineering school students aged from 23 to 26 participated in this experiment. Before the experiment, the LCD used in the experiment needed to warm upat least 30 minutes to obtain a stable color response.Each participant had to judge all the 18 images at 11 levels with 3 brightness levels, which comprised of 594images. They were asked to answer a yes-no question with regard to“do you think this LCD panel will be accepted by customers?” and the acceptance threshold values calculated from these answers were treated as dependent variable. The sequence of viewing images was randomized for all participants.
Fig.5. The transformation model of CNT-BLU to CNT-BLU with LCD screen and transformation model of luminance of LCD to its gray level
IV.RESULTS
The acceptance thresholds of all the 594 images used in this study were calculated from the responses of all the participants. Their luminance uniformitiesof each image for the three methods, namely, VESA Standard, U Formula and Line Uniformity were also calculated. Fig.6 shows the scatter diagrams between VESA Standardand acceptance threshold under three brightness conditions. Sigmoidal modal, which was widely used to describe the threshold of human perception, was used to fit the diagram. The R square values of each modal were from 0.225 to 0.392as shown on Fig.6 Same regression model applied to U Formula showed that the R square values were between less than 0.001 to 0.487 as suggested by Fig.7Finally, it can be seen that Line Uniformity generated the best results from 0.801 to 0.927 (Fig. 8), which means, luminance uniformity that measured using Line Uniformity made a better prediction to human perception. On the contrary, the use of VESA Standard and U Formula could not provide such consistency between human perception and the measured uniformity. R square values in these two cases were under 0.4. However, VESA Standard, that uses only maximal and minimal luminance of measured points had higher R square value than those of U Formula, which requires information with regard to all points on the measured line. Hence, the use of VESA Standard had better efficacy than the use of U Formula.
Fig.6. The acceptance threshold figure under three brightness condition: VESA StandardFig.7. The acceptance threshold figure under three brightness control: U Formula
Fig.8. The acceptance threshold figure under three brightness condition: Line Uniformity
Even the Line Uniformity fitted better against acceptance threshold, the brightness affected the fitting as shown in Fig.9.Given the same calculated Line Uniformity, high brightness generated a better acceptance threshold than lower brightness. To obtain a better predictive model, equation 6 was modified and the luminance used to calculate Line Uniformity was standardized with brightness. The result was shown in Fig.10.Thethree threshold functions for different brightness conditionsoverlapped each other andthe undesired brightness effect almost disappeared. TheR square value for this combined functionwasabout 0.88, which means the luminance uniformity measurement still maintained a good consistency with human perception across different brightness levels.
V.CONCLUSION
The focus of this study was to find a more effective way to measure the uniformity quality of the under-developing CNT-BLB since many current uniformity standards are not suitable for this purpose. Line Uniformity was proposed and compared with another two widely accepted uniformity measures, namely, VESA Standard and U Formula. The results showed that Line Uniformity outperformed VESA Standard and U Formula based on human perceived judgment using a set of simulated CNT-BLU images. Although some may argue that VESA Standard and U Formula are more efficient than Line Uniformity, the calculation for Line Uniformity is rather simple and not seems to be a problem, especially when the production rate is relatively low during the CNT-BLU development stage. The U Formula, which also relies on the information of a whole line, even performed worse than VESA Standard. Even given thesame amount of information like given to Line Uniformity, the calculation from six lines using U Formula only generated an R square value of 0.329, far below the Line Uniformity case. As we mentioned above, CNT-BLU has the typical defect of mottled appearance, the statistics that use in U Formula may be too roughfor describing such detailed variation betweenadjacent pixels. The results suggest that Line Uniformity may be a more appropriate indicator than VESA Standard and U Formula for evaluating the luminance uniformity for CNT-BLU during its developing stage. Since CNT-FED has the same light emission principle with CNT-BLU, we expect that this Line Uniformity can also be applied to CNT-FED, which may need a longer developing time due to its complexity. The different light emission principle of CNT-FED and its mottled background compared with LCD might also suggest that a line-based luminance uniformity measure may be more appropriate than a point-based uniformity measure, not only for its developing stage, but also for commercialized CNT-FED in the market.
Fig.9. The acceptance threshold under three brightness conditions: Line Uniformity
Fig.10. The standardized of three brightness conditions
VI.ACKNOWLEDGEMENT
This research was partially supported by the NationalScience Council of Taiwan, R.O.C. (NSC96-2221-E-035 -023.)
VII.REFEREENCE
[1]Chu, M., Coleman, Z., Henrickson, K., and Yeo, T. (2006). Novel high brightness LED backlight design and optimization. ADEAC Conference.