I was wondering if increased granularity of measurement and calibration of the grayscale would improve performance. For example, when using the Lumagen I have never seen any improvement in going from 12-point to 21-point grayscale.
I decided the give this a comprehensive test. Using the AccuPel 5000 and its associated desktop application I was able to put together a comprehensive suite of test patterns in a few minutes. I included every single 8-bit point from video white down to 10% stimulus. That's digital 235 down to 38, or 197 discreet points, approximately one test pattern for every 0.4% of stimulus. This exceeds by a large margin any level of grayscale granularity offered in any calibration package. The best I know of is the Lumagen's 21-point (5% increments).Methodology
I started by doing a manual 11-point calibration. This allowed me to get all of the measured dE values at 1.4 or below. These calibrated points are shown in red below.
Next, I ran the 197-point suite of AccuPel test patterns, measuring at each level and recording and then exporting the xyY values using ChromaPure's Raw Data module. These measured points are shown in green below.
Finally, I plotted the the red and green points on the same graph to see if the measured values deviated meaningfully from the calibrated points. If they did, then that would show conclusively that an 11-point calibration lacks sufficient granularity and you need to calibrate at more points. If it didn't, then that would show conclusively that going beyond 11-point calibration (at least with the Lumagen) would provide no value.
The results were interesting and not exactly what I expected. Here's the first test.11-Point Calibration
As you can see, down to 20% stimulus the measured points fell--with a very few exceptions--within the desired dE range. This shows that additional calibration points--increased granularity--would have served no purpose. However, it also shows that below 20%, the calibration falls apart. 15 of the 22 measured points between 20% and 10% had dE values in the 3-6 range. This suggests that increased granularity would be useful at the very low end.
Next, I repeated the measurements between 20% and 10%, digital 60-38, after adding an additional calibration point at 15%. Since I had already calibrated and measured the points above 20% and saw no issues, there was no reason to measure and calibrate these again in 5% increments.21-Point Calibration
As you can see, although this improved performance between 20% and 10%--now there are only 10 points above the desired dE range--it still shows that the calibration did not hold very well at the very low end.Conclusions
For the great majority of the video range, calibrating grayscale at 5% increments offers no improvement over 10% increments. However, below 20% additional points of calibration would seem to be worthwhile. Nonetheless, the available 5% increments are just not sufficient in this regard.
If technically feasible at a reasonable level of effort, I would like to see the Lumagen 21-point calibration in 5% increments (in which most of the effort is wasted) replaced with a 15-point suite of readings where 10% increments are used above 20% and 2.5% increments are used from between 20% and 5%. Readings below 5% simply are not practical.
This suite of test patterns and calibration points would run faster than the current 21-point system and offer better results.
I invite anyone who wants to spend about an hour on this to attempt to repeat my results. It is always possible that they are the consequence of some strange characteristic of the display I measured--a Sony LED--but I doubt this. I chose a LED because they offer exceptionally stable light sources.