| VideoToolbox | Psychtoolbox | Eyelink | Tips |
| Home | Intro | Tutorial | Download | Mac OS9  |
OSX  |
Win  |
USB | Questions | Library | Forum | Help |
By Denis Pelli
Welcome to DISPLAY TEST. Here you can determine whether your display is suitable for high-accuracy rendering, for use in vision research. These tests are simple images and animations that you can evaluate visually in your browser, right now. You won't need any equipment except a piece of opaque cardboard with a roughly 1 cm hole in it.
Tests 1-5 (primarily for CRTs) assess pixel independence, which is assumed by the standard CRT model used for display characterization, but is usually not attained by even the best monitors for certain stimuli.
Tests 6-8 (primarily for LCDs) compare the luminance of static and drifting gratings, looking for nonlinear effects of slow LCD switching.
NOTE: The main stumbling block for use of LCD displays for accurate rendering is their very high dependence on viewing angle, which is not assessed here.
The tests are suggested by Brainard,
Pelli, and Robson (2002), but you can do the tests without having read our
chapter.
Brainard, D. H., Pelli, D.G., and Robson, T. (2002). Display
characterization. In the Encylopedia of Imaging Science and Technology.
J. Hornak (ed.), Wiley. 172-188. 
Test 1. Compare the luminance of vertical and horizontal gratings. Limited bandwidth of CRT video amplifiers generally makes the vertical grating (which alternates black and white on successive pixels) dimmer than the horizontal grating (which alternates only once per raster line). The visual contrast threshold is about 1% luminance difference for text or a black-white edge. This test is most rigorous when you back up far enough that your eye cannot resolve the gratings, so any effect of the grating's orientation on luminance must occur in the rendering, not in your visual system. Bear in mind that there are visible edge artifacts (the gratings end on white or black, not gray) that you'll be able to detect even if the luminances of the gratings match. Failures of this test (differences in luminance between vertical and horizontal gratings) are most likely due to insufficient video bandwidth. You should avoid using stimuli that have such high video frequencies (measured along the raster line). To achieve this, you may want to increase the observer's viewing distance.
TRY TEST 1. (Then click your browser's Back button to return here.)
Test 2. This is identical to Test 1, except that the cell size has been doubled, vertically and horizontally. This makes the error a smaller fraction of the cell. Hopefully displays that failed test 1 will pass this test.
Test 3. This is a two-frame animation that repeats endlessly. You will need an opaque piece of cardboard with a roughly one centimeter hole in the middle. Use this occluder to isolate part of the steady white bar in the center of the display, obscuring the flashing surround. You are trying to assess how stable the light is in the central bar. Most displays will fail this test to some degree. Failure due to poor high-voltage regulation will be worst at the bottom of the display and not present at the top. (Does the bottom of the vertical bar jiggle, moving down when the surround goes bright? Poor high-voltage regulation will cause the image to expand when the load (mean luminance) is high, pulling down the voltage, because the electron, traveling more slowly, has more time to be deflected before hitting the faceplate of the CRT.) Failure due to incomplete DC restoration will also be zero at the top and worst at the bottom, but can be distinguished by a further test, below. Failure due to dynamic brightness stabilization (a "feature") will be independent of vertical position.
Test 4. This is a three-frame animation that repeats endlessly. You will need the opaque piece of cardboard with a roughly one centimeter hole in the middle. Use this occluder to isolate part of the steady white bar in the center of the display, obscuring the flashing surround. You are trying to assess how stable the light is in the central bar. This test presents a constant load on the high-voltage power supply, since the surround is always maximum brightness of one gun (red, green, or blue), and the power supply is common to all three guns. Thus poor high-voltage regulation will not affect this test. Any flicker of the central bar must be due to incomplete DC restoration (which will be worst at bottom, absent at top) or dynamic brightness stabilization (which will be independent of vertical position).
Test 5. This is very similar to Test 4. The only difference is that the background successively turns OFF one gun at a time, keeping the other two on. (In test 4 the background turned ON one gun at a time, keeping the other two off.)
Test 6. Compare the luminance of static and drifting gratings, looking for nonlinear effects of slow LCD switching. This test is most rigorous when you back up far enough that your eye cannot resolve the gratings, so any effect of the grating's motion on luminance must occur in the rendering, not in your visual system. Bear in mind that there are visible edge artifacts (the gratings end on white or black, not gray) that you'll be able to detect even if the luminances of the gratings match. Failures of this test (differences in luminance between static and drifting gratings) are due to slow LCD switching. Sorry about the jerkiness; we're using a GIF animation, requesting 17 ms per frame, but it's up to your browser to do it. Quitting other applications may help minimize the interruptions.
TRY TEST 6. (Then click your browser's Back button to return here.)
Test 7. Just like test 6, but the grating duty cycle is reduced from 50% (1/2) to 25% (1/4).
Test 8. Just like test 6, but the grating duty cycle is reduced from 50% (1/2) to 6% (1/16).
visitors since 23 June
2001.
Denis Pelli
psychtoolbox@yahoogroups.com
27 September 2004
