Test Pattern for Wide-screen Displays
Getting the right aspect ratio to your home theater display.
So a couple years after getting a 27" LCD display, I finally got fed up with the ugly stretched image the display showed. There aren’t many options for widescreen source material until we switch to digital cable, so most of the TV we watch is standard definition (SD). If you watch SD, 4:3 video on any display for a few weeks or a month you’ll end up burning-in the center of the screen, leaving the wide left and right margins darker with a very definite demarcation line when you finally start watching widescreen (16:9) content.
Although I stretched the standard TV signal (Comcast Cable, pipped through a Series 2 TiVo) to go ‘widescreen’, Amy and I have gotten used to it for the most part. We just think everyone on TV is about 40 pounds heavier than they are, since they’re stretched left to right about 25% wider than they should be.
What I did to get the aspect ratio correct on my Mac Mini
But I knew I could do better with the signal coming from the Mac Mini. I knew that there were some powerful utilities out there like SwitchResX and DisplayConfigX that could really play around with your Mac’s video output. All I needed was about 2 days of reading web pages, messing around with the settings, and creating this test pattern to make sure I got the aspect ratio right.
The “messing around with settings” part was mostly using VNC to control the Mac Mini from my PowerBook while I watched the results of my experiments on the 27" the Mini was hooked up to. If you’re playing around with video settings like this I cannot emphasize enough how important it is to use something like VNC/Apple Remote Desktop to get things going. Playing with the video settings will screw things up. The 8+ hours it took me to find the right setting would have been 80+ if I had to safe boot and recover from each experiment each time I messed things up.
The last part of the process was creating a test pattern that I could use to figure out how close I was getting to my ideal 16:9 display. The test pattern was made for testing just a few items: the size of the display, the aspect ratio of the display and the gamma or grayscale settings. I’m quite pleased with the result of the test card and wanted to share it with others.
The test pattern image and what each part means
The Square Target: Aspect Ratio
Each area of the test card helps to identify some aspect of the display’s operational characteristics. This middle most portion of the image is meant to appear dead center on the screen. You see the Square Target here that is bisected horizontally and vertically by a 2 px black line. This figure should appear as a perfect square on the screen that it’s displayed on. If you want to be absolutely sure, you could actually break out a ruler and put it up against the screen to make sure the square is actually a square. Layered on top of the square target is a couple of diagonal lines that form an X over the whole image. The arms of the X are 1 pixel wide and will cross over a perfect rectangle that is 16:9, which is the whole reason for this test pattern.
Pixel Counting: Total Screen Size
I’ve marked out a few common sizes for the display’s resolution. Although my version was only trying to target 1280 x 720, I made additional sizes of 1366x768, 1820x1024 and the Full HD size of 1920x1080. I read online that a lot of LCD panels work at the 1366 x 768 size, and the 1920 x 1080 box is the maximum HDTV size.
Multiple Sized Checkerboards: Perceived Resolution
I took advantage of the checkerboard in the lower square to try a few variations of the checkerboard at progressively larger sizes. The checkerboard uses two pixels, three pixels, four pixels and so on up to six pixel squares. This can help you identify how much detail you can see in a display. Once you get a look at the image on one display you can get a feel for the resolution when compared to another display. The checkerboards are also offset by one pixel in each row so as to help find at least one row where the edges should be perfectly sharp.
Horizontal Gradient Check: Gamma Check
This is one of the cooler portions of the test. The top bar is alternating black and white vertical lines that from a distance should appear gray. (If you can still see the vertical lines try to “blur” your vision of the bar to make seem gray. I usually un-focus my view or squint in a way that makes the lines blur together.) Make a note of where the gray becomes roughly equal to the gray in the gradated bar just below it and call that Point A. If the “gray point” or gamma setting is right, the point where the grays are equal (a hypothetical Point B) will be directly below the Point A.
Tick Marks and Vertical Gradient Check: Anti-aliasing
This image shows the gradient bars in a vertical orientation as well as the intersection of the “ticks”. The ticks are a dual set of 1 pixel and 2 pixel lines. The taller set is 2, 4, 6, and 8 pixels wide in each step. The smaller ticks are 1, 2, 3, and 4 pixels wide in each step. This can be used to gauge the blurring from anti-aliasing that the computer is producing under any sort of scaling routine.
Stretched Pixels: Anti-aliasing comparison
This is probably the least useful portion but I wanted to see what was going on with the anti-aliasing of the display versus the anti-aliasing of Photoshop. This section takes a simple 1x1 pixel checkerboard and then stretches it in 10% increments in one row, but compresses it in 10% increments in the other row.
Scoring a 100% perfect
Finally, the real goal of this test pattern is to get a screen that has no moiré pattern when judged against the checkerboard gray of the background of the whole image. If you get this test pattern to display with an even gray checkerboard that has no alternating dark or light patches then your video output is matching the display’s native resolution. That’s pretty much the holy grail of getting this right.
How to Display the Image
I tried a number of different file formats including the obvious ones which should have worked perfectly: TIFF, PNG, BMP and even PICT, but the most accurate rendering of the image comes from using the GIF format. The GIF format is a very simple format as these things go and seems to render more consistently across all the platforms I tried. If you’re interested in the original layered PSD file, I can send it to you. (I didn’t want to post a 13MB image to this web hosting account.)
To display the image I originally just loaded it up in a web browser (Safari was handy) and made sure it was displayed at 100% and not scaled to fit the window. Preview was another option, but by the end of this project I figured out the best way to display the test card was to use it as a desktop background. The key is to make sure it’s displayed as “Centered” rather than the default “Tiled”. This is also why the pixel count rectangles are centered on the card.
It should be obvious by now that this is not a color testing image. If there is a request, I could put one together, but I’m sure there’re other people who could do a better one.
This image also doesn’t take into account non-square pixels that are part of many broadcasting formats (NTSC, PAL, etc.). However, from what I can tell if you can get the desktop of the Mac to display properly, then the software that displays any broadcast signal (DVD Player, Front Row, iTunes, EyeTV, etc.) will automatically take care of that issue.
I am grateful for the following links: Dr. Alvy Ray Smith's Memos (In particular A Pixel is Not a Little Square, a Pixel is Not a Little Square, a Pixel is Not a Little Square! (And a Voxel is Not a Little Cube)), the BBC Test Card Gallery, the HDNet test pattern and the great threads at AVSForum.