Metrology in ISO and VESA Standardization for FPDs

Metrology in ISO and VESA Standardization for FPDs IDIM 2003 Daegu, Korea July 10, 2003 Paul A. Boynton and Edward F. Kelley NIST (Bldg. 225 Rm. A 53) 100 Bureau Dr. , Stop 8114 Gaithersburg, MD 20899 -8114 www. fpd. nist. gov NOTE: Certain commercial equipment, instruments, materials, systems, and trade names are identified in this paper in order to specify or identify technologies adequately. Such identification is not intended to imply recommendation or endorsement by the National Institute of Standards and Technology, nor is it intended to imply that the systems or products identified are necessarily the best available for the purpose.

GOOD DISPLAY METROLOGY—WHAT IS IT? Measurement methods used in standards and industry need to be…` Reproducible — Everybody can get the same results on the same display using appropriate instrumentation. This is probably the most important goal of metrology. Robust — Insensitive to small changes in the measurement apparatus that will affect the ease with which reproducibility is attained. Unambiguous — The method is clearly stated and easily understood. Important details that are required for success are not left out. Extensible — Applicable to as many different technologies as possible permitting intercomparisons of technologies. Distinct — The name of a measurement method must be chosen so that it is not confused with another metric. Honest — The measurement method is not devised to hide an obvious deficiency; redefining familiar terms to cloak a problem.

. . . and we might be tempted to add (sometimes the following are not possible). . . Accommodating — We want to enable as broad a range of apparatus as possible. Accessible — Requiring the use of unusual, highly specialized, or otherwise arcane apparatus or methods would be avoided unless it is necessary (e. g. , for people who influence written standards in order to sell their apparatus when it is not necessary to do so). Simple — Procedures should be made as uncomplicated as possible, avoiding deliberate obscuration from elitism or exploitation (e. g. , deliberately making the standard so difficult to use that only a few experts can use it). Meaningful — Properly captures the visual experience for task and environment. Measuring what the eye appreciates should not be sacrificed for some related esoteric measurement method of limited use. Example: Character contrast of a black letter on a white screen must first be measured correctly accounting for veiling glare in the measurement apparatus (discussed later). (Calculating that contrast on the basis of the Gaussian beam profile of an electron beam in a CRT to avoid measuring the real contrast would not be a measurement of the contrast; besides it would only be applicable to CRTs. ) After a proper contrast measurement is made, if need be, an appropriate vision model can be applied to determine how the eye appreciates that contrast.

Why do we need good display metrology? Level Playing Field — Competition Between FPDs within a technology – Which LCD is better for my purposes? Between different technologies of FPDs – How do we compare a plasma display with a front-projection display (or HMD)? Specification Language Defined Task-dependent specifications possible – How do the contrast requirements differ for a display used in a dark room vs. a display used in a bright surround? Enables clarity and removes ambiguity – Why should we use darkroom contrast vs. ambient contrast vs. highlight contrast? Clear and understandable measurement standards Measuring What the Eye Sees Ergonomics and vision science must be based upon good metrology – “Measure twice, cut once. ” It would be sad to set a compliance standard for a minimum contrast if that contrast was measured incorrectly during the research.

Not So Simple!!! Complications: Photometry-colorimetry uncertainties START at 1 % level. Anticipated luminance uncertainties in the field can be from 2 % to 4 %. Be very happy if you get 1 %! Eye response quasi-logarithmic yet measurement comparisons linearly based, so we don’t see some of the problems we are measuring. Display output can drift in time both in color and luminance.

TASK DEPENDENT SETUP Proper Setup Depends Upon Display Task. How will the display be used? What environment (ambient, surround)? Are there manufacturing setup specifications? Gray scales near black and near white are often useful, but may not be sufficient. [FPDM 301 -3 A] Might also try a face as well as a scene, if appropriate to task. A face may show up problems that are difficult to see with other patterns simply because our vision processing is very sensitive to faces. Available: http: //www. fpdl. nist. gov/patterns. html as NISTSU*. *

Warm-Up Time May Be Needed During the warm-up of the display is a good time to examine the display for defects and problems. Try out many different patterns and images suitable to the intended display task. [FPDM 301 -2 D] For examples of patterns see: http: //www. fpdl. nist. gov/patterns. html Setup Conditions Remaining Fixed During a series of measurements the task-specific setup conditions should not be changed to improve any single measurement, unless the task calls for such changes. [FPDM 3012 E, 305 -3] This has to do with honesty: We want to measure the display the way it will be used. Unless the task calls for tweaking the display for various needs, the controls should be set after the display is fully warmed up and all measurements should be made at that setting.

DIAGNOSTICS Diagnostics—A Wise Choice! Often Ignored Time consuming — slow us down! Ostrich effect — we don’t want to know if we have problems. We can’t afford the extra equipment to do them. Benefits Problems revealed early. Confidence in apparatus, personnel, and procedures. Awareness of how simple things can go wrong. Not a clue!!! Diagnostics!? What do we need diagnostics for? ! We know what we're doing!

Stray Light Within Detector – Veiling Glare Complex Lens Iris CCD with Cover Glass Shutter Photopic Filter Image Object Original Reflection off of internal lens structure Reflection between lens surfaces Veiling Glare Lens Flare

Example: Checkerboard Displayed on Laptop Scientific-grade CCD, TEC, 16 -bit (approx. ), using a good 35 mm camera lens.

Veiling Glare Can Affect Full-Screen Measurements Measurement of Full-Screen White Comparison of two identical luminances having different angular sizes. Same screen with & without mask (1. 5° or 15° vertical angular diameter of white area from lens of detector) 1. 5° Subtense with Mask 15° Subtense without Mask Increase in measured luminance with mask removed: Instrument #1 0. 4 % Instrument #2 1. 3 % Instrument #3 4. 8 %

Veiling Glare Can Affect Box Measurements Measurement of Black Rectangle on White This shows how important it is to anticipate veiling glare in the detection system. Same screen with & without mask (1. 5° mask hole, 15° vertical angular diameter of white area from lens of detector) Increase in measured luminance with mask removed: Instrument #1 50 % Instrument #2 325 % Instrument #3 1200 %

Use of Masks — Flat and Frustum Detector DETECTOR VIEW USING FRUSTUM Frustum aperture Detector aperture Side view with flat mask Side view with frustum mask With some very high contrast displays, the back reflections off flat masks can cause substantial errors.

ROBUSTNESS Three-Component Model of Reflection 1. SPECULAR (producing a distinct virtual image of the source) 2. & 3. DIFFUSE (has TWO components): 2. LAMBERTIAN (like matte paint) & 3. HAZE (fuzzy ball in specular direction) Reflectance can be thought of as having three components: r = r. L + r. S + r. H, where the diffuse reflectance is rd = r. L + r. H.

Three Component Reflection Model, Cont. Specular, Lambertian, Haze Virtual Image (if there is a specular component) Specular Only Haze Only Lambertian Only All Three

Three Component Reflection Model, Cont. BRDF — Three Components: Bidirectional Reflectance Distribution Function A generalization of L = q. E. We will drop the wavelength l and polarization p dependence. B = DL + S + D H d. E, element of illuminance Boring Alert

Three Component Reflection Model, Cont. Like the Lambertian component, the haze is proportional to the illuminance; but like the specular component, it follows the specular direction. The luminance of the specular component remains constant with change in distance. The peak of the haze changes with distance (according to illuminance). Because the haze peak adds to the specular image, it can appear that the specular image is changing its luminance, but that is not the case. Look carefully at the specular image, you will see it is not changing its luminance when you separate the specular image from the haze peak.

Three Component Reflection Model, Cont. Observed Luminance = Lambertian Component + Specular Component Background Distinct gray image + Haze Component Fuzzy ball

Three Component Reflection Model, Cont. Simple BRDF (In Plane) Extremes: q = 5° Shown qi < 0 Detector Lambertian (flat) Specular (spike) Haze is in between. q q Source Haze characteristics: Proportional to illuminance Directed in specular direction qi > 0 FPD Sample SPECIALLY PREPARED SAMPLE (DL + S + DH) 100 Specular Peak Haze 10 Component 10 1 NOTE: 3 to 5 orders of magnitude possible (or more!—your eye has no trouble seeing this range!) BRDF (1/sr ) us bf 0. 1 -10 Haze 1 io at Peak n c O -5 0 5 10 RESOLUTION < 0. 1° Component 0. 1 Lambertian Component 0. 01 -80 -60 -40 -20 0 20 Light Source Angle (degrees) 40 60 80

Three Component Reflection Model, Cont.

Three Component Reflection Model, Cont. Much happens within 1°

Three Component Reflection Model, Cont. Three components in BRDF often seen in CRTs

Three Component Reflection Model, Cont. In the most general case, when there is a Lambertian, specular, and haze component, there at least four parameters that are needed to specify the reflection characteristics since haze has a peak and a width (at the very least). KEY POINT If we only make one simple measurement or two, the problem is underdetermined an Note infinite number of displays can measure the same and look different to the eye! This underscores the need to make several different measurements to adequately characterize reflection from displays.

Three Component Reflection Model, Cont. With Haze, Measurements Can Be Sensitive to the Geometry of the Apparatus. . . Lens diameter Focus Source size Source distance …? Example: ± 1° misalignment of apparatus can result in 30% errors in measured reflected luminance. Haze Reflection Need Not Be Symmetrical. Star patterns and spikes further complicate a full characterization of reflection, accomplished only via a complete BRDF.

Three Component Reflection Model, Cont. HAZE COMPONENT COMPLICATES REFLECTION MEASUREMENTS! Specular and Lambertian components are easy to measure, but whenever haze is non-trivial, the measurement result can become very sensitive to small perturbations in the apparatus, affecting reproducibility and robustness. Diameter of acceptance area of detector can affect luminance measurement. Luminance meters are calibrated against a uniform source (Lambertian).

log 10 BRDF Small Specular Source Angle Relative to Detector View Very sensitive to apparatus settings. Looking in the direction of greatest changes — peak region of BRDF, haze contributions seriously affect results. Not robust, except for simplest reflection properties, i. e. , without haze—even then can have problems.

log 10 BRDF Large Specular Source Angle Relative to Detector View Captures specular, haze peak, and much of haze profile. Fairly robust, but angles may be a problem. Focus on display is sufficient. Works with haze because ‘when one side goes up the other side goes down. ’ Might improve by using a larger source subtense (angular aperture) than 15°.

log 10 BRDF Ring Light Source qr = 20° Angle Relative to Detector View Avoids haze peak and specular. Captures the tail of the haze (depending upon width) and some contributions from the Lambertian. Fairly robust, but distances may be a problem as might measurement field angle. Works with haze because ‘when one side goes up the other goes down. ’ May benefit from using larger source subtense angle.

Used two configurations: Fiber-optic input provides illumination of interior log 10 BRDF Diffuse Source Angle Relative to Detector View Display inside large integrating sphere rd/q = 0. 0557 (± 0. 3 %) over range qd = 6° to qd = 20° Sampling sphere on display surface rd/q = 0. 0554 (± 3 %) — (exit port soft , plastic foam) This is an example of a robust method. Insensitivity to geometry and apparatus! The disadvantage is that it combines all three reflection components together.

Most Robust Apparatus Diffuse Source: VERY ROBUST: Robust in parameter settings AND apparatus employed. Exploits and combines all three reflection components, Lambertian, specular, and haze. Many say, “not practical, ” “not realistic, ”…. NOT TRUE! Not so uncommon after all: Beach on a cloudy day Snow field on a cloudy day Light living room with light furniture and even illumination Bubble helicopter in cloud Horizontally held PDA outside on a cloudy day… People don’t like it because it is tough on their displays, gives low contrast values. Results scalable to many diffuse environmental lighting levels.

Large Specular Source Rather robust. Exploits and combines the peak of the haze component and the specular component as well as adding in some of the Lambertian component. More work is needed to determine if there are better settings (e. g. , larger source subtense) that will secure more robustness—perhaps two settings. A larger subtense approaches diffuse. A smaller subtense captures more of the haze peak and specular with possible loss of robustness. ?

Variable Radius Source Method (VRSM) Isolation of Specular (z) and Haze Peak (h) Method Under Research Apparatus geometry may be important for extracting any haze-peak information. Focusing on source rather than display may be required for specular component measurement. More work needed. Non-smooth surfaces (e. g. , orangepeel like or heatlaminated) or specularonly displays that exhibit spikes may create problems for the method. L z. Ls Can also get haze peak h from curvature near r = 0. r LS r

CANDIDATES FOR ROBUST METHODS Diffuse Source Large Specular Source NOTE: These methods integrate about the normal (f axis) or specular direction thus integrating the effects of complicated haze or specular reflection profiles (spikes and stars, etc. ). Some displays are entering the arena that are not smooth or where all reflection surfaces may not be primarily from a single plane. These methods accommodate such displays. Note

CONCLUDING COMMENTS Intuition (Gut Feelings!)? If you have a gut feeling that something is wrong, it probably is. On the other hand, if everything seems perfect, it probably isn’t.

Look for problems, be suspicious. A bright display can light up a dark room, are you measuring the reflection of your white shirt or the side of a lightly-colored instrument (or wall) along with the screen color? How about equipment lights and displays in the room, do they reflect in the screen being measured? Look and see. Don’t assume. If you can see it, the instrument might be affected by it. Don’t over document and then under document! Don’t spend so much time documenting untested and preliminary apparatus and data so that you can’t finish the measurement—it’s like polishing garbage. Take the time to document thoroughly after it is working properly. P 12 —Lest We Forget Working at the Bench. . . Perpetrating paperwork, poppycock, plus protocol paralyzes promising project progress producing poor products.

Whom do you trust? Don't trust anything or anyone (as much as possible), always try to verify things you are tempted to assume, prove to yourself that everything is working properly and that you are not making inappropriate assumptions.

Trust Your Eye Look at what your instrument is seeing from its perspective. Examples: (1) Eye sees high contrast, instrument measures low contrast—trust your eye. (2) Reflections from nearby objects? (3) Light from instrument indication illumination?

What Is “Good Enough”? We should not compromise good metrology in favor of tradition when that tradition might be based upon inadequate metrology. How was that “limitation” of the eye determined? If the instrumentation used to determine the “rule” is not as good as the eye, then what can’t see, the eye or the instrument? What measurements were made? Was the instrumentation capable of an accurate measurement, how do we know? What diagnostics were performed to prove that the instrument was measuring things correctly? How was “adequately” defined? Be a skeptic!

Comprehensive Document Available VESA FPDM — Flat Panel Display Measurements Standard Features: Specification of good metrology for displays Self-contained measurement procedures Buffet of measurements—use what you need Easy to use and read Extensible—more will be added as needed Adaptable—affords a variety of equipment Accommodating—special needs permitted www. vesa. org Metrology Section, Technical Discussions Section Includes diagnostics, cautions and hints A reasonably priced document ($40) of over 320 pages— VESA 408 -957 -9270

ISO Grand Revision ISO TC 159/SC 4/WG 2 — Visual Display Requirements Under Development Features: Easier to use and read Separate sections for ergonomic requirements, test methods, and compliance routes Specification of good metrology for displays Combines ISO 13406 -2 and 9241 -3, 7 & 8 Scope to be extended to include applications beyond office environment and technologies other than desktop CRTs and LCDs Extendable—more will be added as needed

Courses & Services at NIST Photometry Short Course http: //physics. nist. gov/Divisions/Div 844/facilities/photo/Photometry_Course. html NIST Spectroradiometry Short Course http: //physics. nist. gov/Divisions/Div 844/events/srsc. html NIST Colorimetry of Displays — A Calibration Facility http: //physics. nist. gov/Divisions/Div 844/facilities/photo/Projects/colorimetry_of_displays. htm Based Upon the Four-Color Matrix Method for Correction of Tristimulus Colorimeters NIST Display Metrology Short Course Coming soon (hands on lab work)… if this course goes well… see http: //www. fpdl. nist. gov or http: //www. fpd. nist. gov NIST Flat Panel Display Laboratory — Publications, Links, Overview http: //www. fpdl. nist. gov (“FPDL”) or http: //www. fpd. nist. gov Slides of NIST seminar on metrology: http: //www. fpdl. nist. gov/seminars. html