What to lookout for when calibrating - LightSpace CMS, Calman, and DisplayCal/Argyll
Not all calibration results are made equal, and it helps to know what to look for in the final verification results to ascertain what is good from what is not so good. To help demonstrate this a direct comparison of LightSpace CMS vs. Calman and DisplayCAl/Argyll has been performed, using the exact same configurations of display and probe, with the same patch generator.
All profiling, calibration and verification was performed twice with each system to ensure the results were as accurate as possible.
Note: This is the second Calibration Comparison comparing LightSpace CMS with Calman. The previous comparison, performed with previous versions of both LightSpace CMS and Calman, can be seen HERE.
An outline of the comparison procedure
To perform this comparison, and make it as fair and accurate as possible, a broadcast LCD display was used, with the aim of making things as 'easy' as possible for the calibration procedure. A broadcast display was chosen as many 'home' displays have very poor underlying colour response, so 'stress' accurate calibration, making a direct comparison more likely to show large differences between the systems used.
The approach used for the comparison was based on the 'time taken' to complete the calibration for all three systems - from start to finish. First, a LightSpace CMS 9261 patch sequence profile was performed, which took under 2 & ½ hours (including generation of the calibration LUT), using an i1 Display Pro OEM probe. Calman and and DisplayCa/Argyll were then set to perform 2 & ½ hour profiles, with the exact same i1 Display Pro settings, and without moving the probe from the screen, making the comparison as 'matched' as is possible.
Note: Calman actually took 2 hours 40 minutes, while DisplayCal/Argyll took 3 house 15 minutes to perform the profile and generate the calibration LUT, with around 6000 profile patches each.
The generated LUT from a three systems was applied to the display, and the display re-profiled with the LUT active, using a very large 21^3 verification profile to maximise the verification results.
The display was set to uncalibrated 'RAW' mode, with just black and white set as per the very basic pre-calibration setups listed on the Initial Display Set-up page, just using the BrightnessCal and ContrastCal images - nothing more (no changes to Gamut, Gamma, white point...).
Note: Repeated attempts were made to run a 9261 patch profile with Calman to assess if a greater number of patches (run over a far longer time period than it took LightSpace CMS to run the same number of patches) would help improve the Calman results. Unfortunately Calman crashed repeatedly, and would not complete the profile. Due to time constraints no attempt was made to perform a DisplayCal/Argyll calibration with a larger patch set.
The following images and descriptions show the RAW display (un-calibrated), the final LightSpace CMS calibration verification, and the final Calman and DisplayCal/Argyll calibration verifications. Various points are highlighted to help assess what to look for when assessing the accuracy of a final calibration.
Note: Although the DisplayCal/Argyll calibration was performed twice (as with LightSpace CMS and Calman to verify the results), the results are not as we expected. Normally DispayCal/Argyll LUT calibration is superior to Calman's results - we will re-visit this when time allows.
In the above and below CIE charts the Calman results can be seen to have gamut excursions that go outside the target colour space gamut, specifically around the Green and Red primaries. This means any images seen on the display with this calibration will appear more saturated than they should. And the larger gamut errors occur in the mid-range colours, where they are most visible. The DisplayCal/Argyll results show the gamut to be slightly lower, especially in red. This can be seen in the following animated charts, where the measured gamut triangle is displayed for each stimulus level in turn.
The above animated charts uses Gamut Triangle plotting (which is a feature of LightSpace CMS) and shows the Calman calibration gamut moves beyond the target colour space boundaries, as well as track along the red and green gamut edges, while the LightSpace CMS calibration stays within the target gamut triangle, as well as not tracking along the red and green gamut edges. The DisplayCal/Argyll gamut plots show red gamut is initially low. While these errors may appear to be small, they will be visible.
The above graphs show the original display gamma, and the calibrated results. The results are easier to see in the following Differential Gamma charts.
These results really speak for themselves - the closer to the desired target gamma, the better the result.
As with the Gamma charts, RGB Separation is easy to read. The closer the individual RGB plots overlap each other, and the closer those plots are to the ideal target (the black line), the better the calibration accuracy, assuming the display's native gamut is larger than the target gamut, as is the case with this display. The poorer RGB Separation result with Calman, and to a lesser extent with DisplayCal/Argyll, are fairly obvious.
As with the Gamma and RGB Separation charts, RGB Balance is also easy to read. The closer the individual RGB plots overlap each other, and the closer those plots are to the ideal target (the black line) the better.
RGB Balance is the same as Grey Scale/White Point, and defines the purity of the display's grey scale, so it is one of the more important items to get as accurate as possible. The one issue to understand though, is that when you get towards the 'blacks' (the dark shadows) the colour is predominantly defined by the back-light colour temperature, and cannot be corrected, without 'lifting' the black, and making it 'grey', which is totally undesirable. Additionally, attempting to correct this area of a display with a poor back light will introduce colour artefacts as the calibration is attempting to correct what can't be corrected.
Again, the poorer result with Calman and DisplayCal/Argyll are fairly obvious to see.
Grey Scale Delta-E
The Grey Scale Delta-E is an alternative view of the effects of Gamma and RGB Balance combined, and shows the expected visual error plotted as a Delta-E graph. The black line at 2.3 is the JND level (Just Noticeable Difference), and theoretically any error below that 'should' be visually invisible. The reality is different, and maximum Delta-E values closer to 1.0 are more desirable.
The Delta Distribution graph is a quick way to assess the overall calibration accuracy for the whole display. Ideally, the graph plots should be as far to the left as possible, and as tall as possible (basically, the taller and narrower the plot, the better, so long as it is all the way to the left of the chart!). Looking at the above graphs the difference between the LightSpace CMS, Calman and DisplayCal/Argyll calibration results are obvious, with LightSpace CMS showing a significantly greater level of overall accuracy.
(Shading has been added to the graph to aid visualisation)
Another way to visualise the difference in overall accuracy is to use the Delta-E range selection drop-down menu, and cycle through the whole Delta-E range in 0.5 step increments. This very visually shows where the errors are. Ideally all errors should be in the first and second steps, as that means there would be no errors above 1.0 Delta-E. While that is near impossible on any display, the LightSpace CMS results do show a very good level of overall calibration, while Calman and DisplayCal/Argyll show a greater range of Delta-E errors.
As can be seen, LightSpace shows most Delta-E errors to be between 0.0-1.5, with no errors above 2.5-3.0, while Calman and DisplayCal/Argyll show most Delta-E errors to be between 0.5-2.5, with errors all the way to greater than 5.0.
Note: This calibration comparison was performed with the newer i1 Display Pro OEM RevB, which has better capabilities than the older original i1 Display Pro, as used for the previous Calibration Comparison, and in conjunction with changes made to the LightSpace CMS 'Colour Engine' has resulted in the far more even Delta-E distribution as seen here.
The PDF reports exported from LightSpace CMS can be downloaded here:
Previous Calibration Comparison
A previous Calibration Comparison using older versions of LightSpace CMS and Calamn can be seen here: