Using Probes for Display Profiling
Performing display characterisation, or profiling, requires the use of a probe to measure and report values from the given display when stimulated by known colour values.
One of the biggest problems with calibration is that not all probes are the same, and it can be very important in gaining the best calibration result to understand the various limits of any given probe, and how best to manage those limitations.
Getting the most from Profiling
There are two main variations between different probes and the techniques they use to perform their measurement functions - the first is colour accuracy for a given display technology, and second the probe's low-light capability, specifically when speed of measurement is compared.
Colour accuracy often has a direct relationship to price, and the more expensive probes tend to use better technology for colour measurement, which often, though not always, means the probe being a true spectroradiometer or spectrophotometer, rather than a filter based colourimeter.
In basic terms, filter based colourimeters rely on the spectral sensitivity of the red, green, and blue filters used to be accurately matched to the expected light source, combined with probe offset settings for different display types as well overall accuracy. The use of red, green, blue filters limits the range of light wavelengths measured, so tend to be less accurate overall - especially over a range of different display technologies. The filters also tend to age over time, so need to be checked and the probe re-calibrated, having it's internal offset values updated to maintain accuracy.
Spectrophotometer on the other hand measures the absolute spectral radiation directly across a wide spectrum, at various frequencies. These multiple measurement bands are combined to provide individual colour measurements. Measuring multiple spectral bands in this way can help improve accuracy greatly.
Low-light accuracy can also be related to probe cost, but in a different way...
The wide spectrum approach of spectrophotometer can mean that low-light readings can be rather slow, or inaccurate, or both, when compared to colourimeters. Dark calibration offsets are also required to maintain accuracy, often via a mechanical shutter in higher-end probes.
Colourimeters, especially at the higher price points, can often be more accurate, or faster and more stable, when measuring low-light levels, due to the simpler direct red, green, blue spectral measurements. So the top-end colourimeters can be preferable to spectrophotometers.
Probe Low-light Specifications
Note: Luma only - Chroma patches will require greater sensitivity
- CR-100: 0.0007 Nits
- CR-250: 0.17 Nits
- CS-200: 0.01 Nits
- CA-310: 0.005 Nits
- DISCUS: 0.05 Nits
- Hubble: 0.034 Nits
- i1 Display Pro: 0.1 Nits
- i1 Pro 2: 0.2 Nits
- Jeti 1211: 0.1 Nits
- K10-A: 0.0001 Nits
- PR-655: 0.7 Nits
- PR-670: 0.035 Nits
The Best Solution?
To gain the most accurate calibration the solution is really to go for the best possible probe... but, for many this cost is overkill, as the top-end probes, especially spectrophotometers, can be 10's of thousands of pounds in cost, and with the tools provided within LightSpace can often not be necessary.
LightSpace provides a range of tools to help overcome the issues associated with lower-cost probes, and can provide results that are very accurate, taking into account some fundamental understandings of the needs of calibration.
The aim of any calibration is to maximise the accuracy of the display within the limits of the display's capabilities, with a focus on the central core of colours within the display's required colour space. Understanding this is very important, as many calibration systems focus on the extreme colour space colours (the primary colours) only. In the real world we see very few 'primary' colours, and focussing on them for calibration can often lead to very inaccurate final results, especially as many displays struggle with accurate primaries.
Additionally, calibrating at low-light levels can be very difficult with the more cost-effective probes available, with the probes introducing errors into the profile that do not accurately represent the display. Combine this with cheap displays that have a colour cast within their back-light, that shows in low-light conditions, but is overcome by the display as brightness is increased. Such back-light cast colours cannot be 'colour corrected' out, although the profile data will cause many colour calibration system to attempt to do so, causing inaccurate calibration results.
A good calibration system will provide ways to overcome such display and probe problems.
Probe Matching to lessen Colour Accuracy Errors
With the desire to have fast profiling, low cost probes, and colour accuracy being a bit of an oxymoron, alternatives are needed, and with colour accuracy and speed there is a fairly easy solution by using a colourimeter based probe for speed, but calibrating or matching it to a spectrophotometer on the same display.
In this way the colourimeter is accurately matched to the spectrophotometer for the specific display to be profiled, producing more accurate results.
The following image shows an i1 Display Pro (i1D3) matched to a CR-250RH Spectro for a display called 'main-1'.
Probe Matching Process
- Attach the first probe, selecting standard probe parameters as required
(You can use any of the default pre-sets, nominally the one that is the closest match you your display type)
- If using the LightSpace inbuilt patch generator, enter a matrix name and press 'OK'
(Use a name that includes the display and probe details)
- Place the probe on the patch window and press 'Measure'
- The patch window will cycle R, G, B and W patches, and save the probe/display matrix data
- Change the probe to the second (Spectro) probe, and repeat the process
(Be aware Spectro's do not use matrix pre-sets)
Note: the order the probes are 'measured' in is not important, and you can measure the Spectro first, with the Tri-stimulus second.
- If using a separate patch window, not controlled via LightSpace, use the RGBW 'Update' buttons in-turn
(Each patch colour needs to be a value of 240 to match the LightSpace patches)
- Enter a name for matrix when prompted
(Use a name that includes the display and probe details)
- With the probe placed on a patch of the matching colour press 'Measure' to take a measurement
- When White is measured the Luma value will be updated too
- Alternatively, manually enter the xy values, remembering the Luma value too when reading White
(As with direct LightSpace matching, the colour patches MUST be based on 240 data, NOT 255)
- Select the Reference Probe/Display matrix from the lower drop-down menu
- Select the Active (in-use) Probe/Display matrix from the upper drop-down menu
- All measurements will now be 'corrected' using the Probe/Display matching function
(You must always use the same Tri-stimulus probe's preset matrix as used when performing the probe matching!)
Using the above probe matching the speed improvements of a colourimeter can be used, and more importantly the lower-light reading capabilities can be employed, while enjoying the accuracy of the spectrophotometer.
The following image shows potential problems with a Calibration LUT generated from a profile using a probe that has struggled with low-level measurements.
Sometimes such errors can be inherent in the display, especially if the display is a cheaper one, where the back-light has an obvious colour cast that the display struggles to overcome as brightness increases. For example, many cheaper LCD displays have a very blue back-light, which can't be overcome by calibration as it is inherent in the back-light. But, the calibration profile and resulting LUT won't know this, and will try to counter the effect. This can cause real issues with calibration in the low-light range.
As the above LUT image shows there are 'kinks' in the shadow area that may introduce unwanted colour artefacts into the final calibration. Using the LUT Manipulation tools with LightSpace CMS it is very easy to remove these errors, while keeping the underlying profile data. The following image shows the same LUT after the LUT Manipulation 'Relax' filter has been used to reduce the error introduced into the calibration by the probe's inability to accurately measure low-light levels.
There are many such tools built into LightSpace CMS to help in generating the best possible final calibration result. See LUT Manipulation Filters for additional LUT manipulation information on dealing with potential black/shadow issues.
What Can't Be Corrected
One of the issues with display calibration that is often overlooked is that if the underlying display has a colour gamut that is below that of the target gamut the calibration can only effect the area of the display gamut that is capable of matching that of the target.
Display calibration systems that major on the primary colours (which in this instance includes secondary colours, as we are talking about the colours at the edge of the target colour gamut) cannot accurately re-map the interior gamut areas that can actually be calibrated.
The result is that it is key that any calibration system understands what can, and what can't be calibrated, and does the best it can to calibrate what can, and doesn't allow what can't be calibrated to adversely affect the final result.
The above CIE diagram shows the profile results from the display referenced above. As can be seen, Green is out of gamut, and the wrong hue (too cyan), while red is slightly out of gamut, and also of the wrong hue (too yellow), and blue is roughly on gamut, but also the wrong hue (again too cyan). White is off axis by being too yellow.
To correct these issues, as much as is possible, the LUT generated from the profile must work to correct the white point, as well as the areas of the display gamut that can be corrected.
The following three cube images show the calibration process for the display referenced above.
The first cube shows the target calibration. Effectively, what a perfectly calibrated display would look like when compared to the target colour space and gamma.
The second cube shows the actual display profile compared to the first cube. Here it can bee seen there is too much green gamut, but not enough blue, as shown in the previous CIE diagram, and red has a non-linear distortion as can be seen in the right hand corner of the cube.
The final cube shows the results of the calibration, and shows that the centre section of the calibration is very close to accurate, while the extremities of the gamut are as close as can be for the limited original gamut of the display.
The above CIE diagram shows the same display re-profiled with the calibration LUT applied (the same as cube 3 above). As can be seen, the calibration is now accurate where it can be, with similar errors where the original display gamut just can't match the target.
This is a good example as to why relying on 'Red, Green, Blue' for gamut calibration checking can be wrong. If the display can't reach those extreme values, trying to force the issue can cause the rest of the display to be grossly inaccurate.
For information on how to verify Calibration via the use of the 'Active LUT' function, see the Active LUT section within the Profiling Manual.
Display Backlight Issues
A good example of display issues that can't be corrected, but cause many colour calibration programs to fail are backlight issues.
An ideal display would have a totally neutral backlight, meaning that black is black. However, many displays, including those from some of the professional manufacturers, have very distinctive colour casts to their backlights. As any backlight colour cast cannot be corrected it is very important that the calibration system understand this, and calibrates the display accordingly.
For example, the above display profiles shows that the backlight has a distinctive blue cast to it, as can be seen below with the addition of 'trend lines' to show the blue backlight bias.
Backlight bias trend lines
With such a bias in the backlight it is imperative for clean calibration that the calibration software 'works with' this bias, and doesn't attempt to 'calibrate it out' as this is both impossible, and will result in unacceptable distortions in the shadow areas of any displayed image.
When calibrating such a display the best calibration will be one that 'works with' the backlight. As stated previously, any calibration system that attempts to counter the backlight will produce very unacceptable results.