Almost all display measurements of performance are done in a dark room. So how do you measure the performance of a reflective or transflective display? Does the performance of reflective displays really match up with the traditional ways to measure them? Qualcomm, the maker of the reflective mirasol display, thinks not and has proposed a test method that addresses one aspect of this issue: color depth.

Matt Brennesholtz
Insight Media Analyst

The color mirasol displays have 2 bits of direct drive color depth for each of the 3 color channels (RGB), plus six additional bits per channel from dithering. If you include only the direct drive colors, this display would be able to display 64 colors while if you include the full 8 bits/color, the display could show a full 16M colors. Neither of these numbers fairly represents the mirasol display. 16M colors would imply full HDTV colorimetry. I saw the mirasol 2.2" demonstration display at SID and while the colors were good, they weren’t that good. Nor did it have 64 colors either, it wasn’t that bad. I took the photo of the prototype 2.2" mirasol display using the ambient front light provided by Qualcomm. If you blow up the photo you can see the spatial dithering in the image. But when I stood there watching still and moving images on the display, it looked pretty good. While the colors were not HDTV-grade, they were certainly better than you would see on a typical cell phone in the same ambient light.

Qualcomm has a proposed solution to this, of course. At SID they distributed a white paper on a proposed Mobile Color Depth test method. This test method is a 5-step approach:

Step One: The display manufacturer’s specifications for addressable bit depth and gamma curve are obtained.

Step Two: The spectral characteristics of the display’s emissive primary colors (if present) are measured, along with the display’s reflectance of ambient light. The display’s white state, black state and primary color states (typically red, green, and blue) are independently measured for both emissive and reflective color.

Step Three: Using the published addressable level and gamma curve specifications together with the white, black and primary color measurements, a luminance versus addressed-level plot is made. This should include both direct-drive and dithered levels. If spatial or temporal dithering is done, the S-CIELAB color appearance model is used to remove any color level that does not match the target primary color.

Step Four: Calculate the number of levels that are separated by about 1 Just Noticeable Difference (JND) using the S-CIELAB JND formula. The number of levels counted will typically be much fewer than the numbers calculated by the formula: Levels = 2^(Bit Depth)

Step Five: Repeat steps Three and Four for each primary color. By multiplying the number of levels for the three primary colors together, you get a single number quantifying the number of perceivable colors in the display.

The Qualcomm white paper then applied this process to mirasol displays. For the prototype 384 x 288, 2.2" color display, Qualcomm calculated there were 45 red, 73 green and 19 blue levels, for a total of 45×73x19 = 62,415 colors. This number seems to me to be a better number than the traditional numbers of either 64 or 16M. Note this method counts the number of "Colors" in a generic sense: it actually counts the number of perceivable color/luminance combinations a display can produce.

My one problem with this proposal is in Step 3. In a dithered display is would be difficult even for the display designer with full details of the dithering algorithm to evaluate which dithered states differ noticeably from the primary color. For a competitor’s display, it would be virtually impossible to do this.

Another problem is more political than technical. Why is the proposal limited to mobile displays? The test method could be used on any display to determine the number of perceivable colors the display can produce. After all is said and done, 16M colors isn’t a good number for a HDTV display since the human eye simply cannot perceive that many different color/luminance combinations. When M. R. Pointer calculated in 1998 the maximum number of perceivable colors using L*a*b* and a procedure similar to Qualcomm’s, he got 2.3M for the entire CIE gamut, not just the colors reproducible on a display.

Even if this display metric is not perfect, it shows the display manufacturers are thinking about the problem of evaluating displays for use in non-traditional locations. This is certainly a step in the right direction and I hope to see other test-methods addressing other display properties emerging in the future and working their way through the long and often tedious process of becoming an official measurement standard.