Considering Color Performance in Curved OLED TVs

Considering Color Performance in Curved OLED TVs

One of the creators of the IDMS (Information Display Measurements Standard) takes the measure of the very latest large, curved OLED TVs.  In this first article in a series, he looks at color performance.

by Edward F. Kelley

LATE LAST SUMMER, 55-in. curved organic light-emitting-diode (OLED) TVs from both LG and Samsung were introduced to the U.S. marketplace.  Commercially available large OLED TVs were novel to begin with; the addition of the curved form factor, designed to enable a more immersive viewing experience, made them even more so.  The display community at large has no prior experience with these unique panels, and, therefore, this article describes the innovative work of investigating their performance and fine-tuning the measurement methods needed.

This article compares the color performance of two curved OLED TVs that have recently gone on sale in the U.S.: the LG 55EA9800 and the Samsung KN55S9CAFXZA (see Fig. 1).  Both have 55-in AMOLED displays.  Only one display from each manufacturer was examined, so the data presented here may not be representative of a statistical sampling of such displays (at about $10K retail price per TV at the time of this writing, this may be understood by the reader).1  In this article, we evaluate and compare the sets’ color gamuts under various conditions, viewing-angle properties, and other straightforward characteristics.  In future articles, we will examine a variety of other characteristics.


Fig. 1:  The LG 55EA9800 and the Samsung KN55S9CAFXZA 55-in. curved OLED TVs were tested for this article.  The LG is at the left and is mounted for measurement.  A photograph is shown on the displays for visual interest, but patterns were used for measurements.


The Samsung OLED employs red-green-blue (RGB) subpixels in a horizontal configuration, and the LG OLED uses RGBW (W for white) subpixels in a vertical configuration.  The LG subpixels are white OLEDs covered with colored filters, and the Samsung subpixels are OLEDs that are tailored to emit the specific RGB colors (see examples of the subpixel configurations and spectra in Fig. 2).


Fig. 2:  Subpixel configurations and spectra appear for both tested units.  The Samsung OLED RGB subpixel arrangement is shown in (a), where the subpixels each have two parts separated by a narrow line.  The smearing between the subpixels in the black region is apparently caused by a lenticular treatment (b) covering the pixel surface.  The LG subpixel arrangement is shown in (c), where the main gray scale is produced by the white subpixel and the colored RGB subpixels regulate the color; several pixels of various colors are shown (a pixel is composed of RGBW subpixels in the order of GBRW from left to right).


Modern televisions can be operated in a number of different modes with numerous settings.  All factory settings were used as the basis for these tests.  In both cases, the overscan feature was turned off.  We noted that both displays exhibited a behavior in which the image luminance would slowly dim when displaying a static pattern.  We can speculate that this might be intended to protect the display from static image burn-in, but that was not confirmed.  Unfortunately, this feature made it harder to collect consistent performance measurements based on static images.  The Samsung display had an option to disable the feature; the LG display did not.  This made it necessary to obtain the measurement result before the dimming occurred if at all possible.  A previous review of the LG OLED display has been made by Dr. Ray Soneira and is available on the Internet.2  For the record, we obtained results similar to Soneira’s here.

The Measurement Environment

The displays were measured in a darkroom with no ambient illumination.  The alignment of the center horizontal normal of the curved displays was assured by a laser alignment system to less than 0.25° along the measurement optical axis.  Both displays were tilted back approximately 5° or 6° so that their central normals were pointing slightly upward from the horizontal plane.  All measurements were made at screen center in the horizontal plane below the central normal – we assumed this was the design viewing direction.  A spectroradiometer was used for all measurements reported here.3  We make reference to the new IDMS document that specifies numerous measurements and recommends various practices.4  A stray-light-elimination tube (SLET) was used for all measurements (see ICDM §5.1 p. 47 and Appendix A2.1).  A SLET helps avoid stray-light contamination from bright areas when measuring dark levels.

Zero-Luminance Blacks

One of the most refreshing features of these OLED displays is their ability to show an absolute black level – a zero-luminance black.  Various people have commented that such a black is not necessary for “normal” viewing, but what is “normal” viewing?  When we see a movie with a sparse star field just before a spaceship enters the view, it is wonderful to see a true black between the stars rather than a dark gray with the surrounding bezel blacker than the blackness of space.  Now we can see black as we would if we were in space.  Theme parks may want a black screen so people won’t see the display in the dark tunnel until it flashes imagery designed to scare the riders.  Gamers may want to feel like they are looking out into space from their spaceship, etc.

All modes of these two OLED displays, save one, could show zero-luminance blacks.  They could also show gray levels as low as level 1 or level 2 out of 255 levels above level 0 (black).  The LG display in the THX Cinema mode can show a non-black background almost as dim as level 1 of 255 with a mottled appearance.  This non-zero black for the THX Cinema mode, the apparent result of a marketing decision, is available only on U.S. models.  However, all other LG modes show a zero-luminance black.  If the non-black LG THX Cinema mode annoys a viewer, it is a simple matter to configure one of the two available extra Expert modes to appear like the THX Cinema mode but with a zero-luminance black.  Should it be necessary, the black levels are generally adjustable in both displays to account for some ambient illumination.  Because the black of the LG THX Cinema mode has a mottled appearance and is very low in luminance – of order 0.001 cd/m2 depending upon where it is measured – we list its black luminance as “not measurable” rather than providing a possibly misleading number for the contrast.

Zero-luminance blacks make the display have an infinite contrast no matter what white level is produced (contrast is C = CW/CK with a zero in the denominator).  This is why the IDMS requires the use of the term “undefined” for such contrasts and suggests that the black and white levels be reported separately (see IDMS pp. 46 and 47).  In the display industry, we seem to have reached the point where contrast (luminance ratio of white to black) has lost its meaning and quoting the white and black levels is much more relevant.


OLED displays are a current-driven pixel technology, and, like plasma displays, will exhibit loading characteristics.  Figure 3 illustrates the loading of these OLED displays where a white rectangle starting at 2% of the linear dimensions of the screen expands to full screen.  Note how the white luminance decreases with the increasing size of the white area; this is loading – which also occurs with saturated colors.  Loading may be considered by some to be an undesirable characteristic, but it may not be noticed by most viewers, and much of the television programming does not load the screen as much as a large white or colored area.  Without loading, the power requirements for these displays would be much greater.  Because of this loading phenomenon, a series of patterns that kept the average pixel level (APL) of the pattern a constant was needed in order to test the displays.  Here, we use the term APL to refer to the pre-gamma input signal average (this is not the average luminance level).  We selected a 17-level stepped gray-scale pattern of concentric overlapping rectangles with each gray level (including white) occupying the same size area on the screen and where the black level cycled in exchange with the center rectangle through the other 16 levels.  The overall pattern was 33% of the size of the screen, and the APL of the rectangular pattern area was 50%, resulting in an overall APL for the entire screen of 5.43% (see Fig. 4).  Unfortunately, even the 33%-sized pattern still caused loading effects in some modes.


Fig. 3:  Loading characteristics are tracked for the LG (left) and Samsung (right) units in Standard Mode.



Fig. 4:  The concentric 17-level 50% APL pattern covered 33% of the screen, giving an overall APL for the entire screen of 5.43%.


General Characteristics

Table 1 shows the general characteristics of the tested OLED displays as viewed from the design viewing direction (below the tilted screen normal at the center of the screen),  (see Fig. 5).  We refer to the signal gamut as the sRGB gamut.5  We follow the suggestions of the ICDM §5.1 in not attempting to report infinite or unmeasurable contrasts.  The uncertainties are the typical uncertainties obtained from a quality spectroradiometer.  The spectroradiometer was tested against red, green, blue, and violet lasers prior to these measurements to assure that their chromaticity values appeared on the spectrum locus (i.e., very close within expected uncertainties).  The wavelength locations of mercury lines were also checked and found to be within 1 nm of their published values.


Table 1:  The above main modes used factory settings (except as noted in the main text).  Measurements were made from the horizontal viewing angle in the horizontal plane of center.  The pattern used as a 17-level 50% APL covering 33% of the screen with an overall APL for the entire screen of 5.43% (see Fig. 4).
Display and Mode LW (cd/m2) LK (cd/m2) Contrasta C = LW/LK Relative Gamut Sizeb (c.f., sRGB) Average Gamut Sizecuʹ,vʹ) CCT (K) Gammad (γ) Full-screen K or W Power, P (W)
Samsung OLED               PK = 115
Dynamic 489 0 Undefined 116 0.0197 12 580 Loading PW = 252
Standard 398 0 Undefined 120 0.0159 8945 Loading PW = 219
Relax 358 0 Undefined 120 0.0100 6530 2.58 PW = 215
Movie 221 0 Undefined 104 0.0040 6277 2.27 PW = 188
LG OLEDe               PK = 51
Vivid 373 0 Undefined 118 0.0024 11 030 Loading PW = 213
Standard 389 0 Undefined 118 0.0023 10 230 Loading PW = 210
Eco 267 0 Undefined 118 0.0026 10 010 Loading PW = 171
Game 235 0 Undefined 118 0.0024 9221 2.34 PW = 162
THX Cinema (not adjustable) 130 ~<0.001 Not measurable 101 0.0034 6433 2.20 PW = 95

aSee ICDM §5.1 for a discussion of how to deal with low-level or zero-luminance blacks.
bSee ICDM §5.18.1 Relative Gamut Area for using the sRGB gamut in the (u',v') chromaticity diagram.
cΔ(u',v') between the centers of the sRGB gamut and display gamut.
dGamma values obtained using the gain-offset gamma-offset (GOGO) model (see ICDM §6.5).  A “Loading” entry means that the gamma model is not appropriate.
eThe two Expert modes are not measured or listed because their default factory settings are very similar to the THX Cinema mode and can be completely configured as the viewer wishes as well as allow for zero-luminance blacks.


Fig. 5:  The design viewing direction is shown in the horizontal plane at the center screen with the curved display tilted back.


For both displays, the observed loading effects created a transfer function [L(V)] that could not be easily fitted with a typical characteristic gamma curve as illustrated in Fig. 6.  The gray-scale curve can show a positive second derivative as the luminance increases when loading occurs.  Thus, in Table 1, gamma values could only be reported for the non-loading gray scales for both displays in their dimmer modes. Figure 7 shows how closely both displays matched the sRGB signal and gamut for the 17 gray levels (the white point) and four levels of color above black for the theater modes of each display.6  Near the center of the gamuts were two plus signs that indicated the gamut shift.  These marked the center of the sRGB gamut and the center of the display RGB gamut.  The metric for relative gamut size gave no indication of the amount of overlap of the two gamuts.  The gamut shift metric (in Δu',v') provided some indication of overlap, but not rotation.


Fig. 6:  Fitting a gray scale with loading will not work properly (left) whereas modes with less luminance and without loading can be properly fit with a gamma model (right).  The Samsung display is shown here; the same loading distortion of the gamma curve occurred for both displays with the brighter modes.


Fig. 7:  Accurate gamuts and very small gamut shifts for theater modes were detected at the design viewing direction.


Figure 8(a) shows how the gamuts changed for each display in their Standard modes; the LG exhibited an almost constant gamut with loading and the Samsung exhibited an increased gamut with loading.  Figure 8(b) shows the change in the gamuts for the displays in their theater modes with a change in level.  The LG display did not change very much as the RGB levels changed, but the Samsung gamut increased as the RGB levels decreased.  In terms of color accuracy and reproduction of the input signal, it is better to have the gamut not change with changes in level, loading, or anything else.  However, because people tend to like more saturated colors, how objectionable gamut increases are will probably depend upon whether the viewer is a critical observer or not.


Fig. 8:  Changes in relative gamut size were apparent with loading (a) for Standard modes and with level change (b) for theater modes.


Viewing-Angle Properties

A curved display has slightly different properties when it comes to viewing angle:  For a single observer, it is optimal for a certain viewing distance and the viewing-angle performance is not as significant.  However, for multiple off-axis observers, the curvature could make the viewing-angle requirements more severe because the angles become larger than at the screen center on the side of the screen where the off-axis viewers are located (see Fig. 5).

For this evaluation, a new way to visualize the viewing-angle performance was employed.  This method considered the shift of relative color-gamut area and may eventually be considered for the next major release of the ICDM (ICDM2, see ICDM §9.8 Viewing Angle Relative Gamut Area).  The relative gamut area as a function of viewing angle is graphically shown in Fig. 9.7  That figure shows results out to 60°, but, realistically, most viewers would only use such a curved display out to 45° and probably considerably less than that.  A larger gamut shift as well as a reduction in gamut size can be observed at larger angles for the Samsung display (with a very small shift at ±15°), whereas for the LG display the gamut increases for larger angles but the gamut shift remains relatively constant.  The change in luminance with viewing angle is shown in Fig. 10.


Fig. 9:  A relative gamut area and shift as a function of horizontal viewing angle appears at left and one example of a gamut reduction below 100% with a shift toward the green is at right.  The shift is represented by two plus signs near the center of the gamut triangles.  The separation of the markers indicates the size of the shift in the display gamut compared to the sRGB gamut.


Fig. 10:  The viewing-angle luminance of white is shown in “movie” modes for both units.


It is important to recognize the scale of these (uʹ,vʹ) chromaticity diagrams.  An expanded view of the Movie and THX Cinema modes appears in Fig. 11.  The error bars on the sRGB white point represent the approximate detection limit of a color change in adjoining color blocks – a Euclidian distance of 0.004 on each side of the white point [see ICDM Appendix B1.2 Colorimetry, p. 471; some refer to this as a just-noticeable-difference (JND) for colors].  Ten times those error bars would represent the color discrimination for widely separated colors such as on two different displays.  However, as gray levels darken, our ability to distinguish color diminishes, and the deviations from the white-point color are not as serious as they may appear on these (uʹ,vʹ) graphs.  (A better metric may be the C* metric, comparable to the ΔE* metric, which indicates the distance from the white-color vertical line in the CIELAB color space where the darker colors are less discernible: C* < 2 for the LG display and C* < 3.5 for the Samsung display.)


Fig. 11:  The magnified area of the white point for the theater modes is represented.  The error bars around the sRGB white point represent ±0.004 on each side of the white point.


All these results show that these TVs, especially in their theater modes viewed from the frontal direction, are almost “perfect” with respect to luminance, color gamut, gamma, and black levels, as Dr. Soneira describes.2  However, keep in mind that once these displays are connected to the Internet and upgraded software is downloaded and installed, these characteristics could very well change.

Future articles will address other performance characteristics.  In the meantime, if the hefty price tags do not dissuade you, either of these TVs can offer an unparalleled viewing experience in terms of  gray and color scales, accuracy in color, and especially the wonderful absolute blacks.


1This work was partially funded by LG Display Co., Ltd.  The company’s contribution to this effort is gratefully acknowledged.

2Dr. Soneira reviews the LG OLED display and compares it with plasma and liquid-crystal displays.  See: http://www.displaymate. com/LG_OLED_TV_ShootOut_1.htm.  He uses the term “perfect” to describe several of its modes of operation.  This same fine behavior is shown in both displays in this article.

3Disclaimer: The apparatus described herein are identified only for the purpose of complete technical description: The signals are provided from a computer using an NVIDIA GeForce GTX 570 board with an HDMI (high-definition multimedia interface) output.  The signal output quality is checked using computer monitors to assure that what is delivered to the TVs is correct without artifacts.  The spectroradiometric measurements are made with a Photo Research PR-730 spectroradiometer.  All measurements are made in a quality darkroom.

4Information Display Measurements Standard (IDMS), prepared by the International Display Metrology Committee of the Society of Information display.  The PDF version is available without charge; see,

5The sRGB signal requirements are specified in ITU-R BT.709-5: Parameter values for the HDTV standards for production and international programme exchange, April, 2002.

6These are the harmonized gray levels described in ICDM §A12.1.1 (the appendix) Table 1.  The 17 gray levels are: 0, 15, 31, 47, 63, 79, 95, 111, 127, 143, 159, 175, 191, 207, 223, 239, 255; and the RGB gamut colors were measured on a five-level gray scale: 0, 63, 127, 191, 255.

7The idea for the graphical gamut-area representation has been proposed by LG Display. •


Ed Kelley is a consulting physicist with Keltek LLC in Longmont, Colorado.  He can be reached at