Error!

Spotting and understanding errors with profiling/calibration workflows.

Many issues faced when attempting to calibrate a given display are often not LightSpace associated issues, but can be difficult to assess without some understanding of what is actually expected, and what could be causing the issues.

This page will define possible issues, and their likely causes.


LUT Calibration & Verification Issues

LUT Verification

The first issue is how to verify the accuracy of any calibration, as this can often be difficult to understand, and the results can sometimes suggest that the LightSpace result is not as good as it should be, even if that is not actually the case.

As verification is the only way to assess the accuracy of any calibration it is a critical process that needs to be fully understood, especially when the reported verification results show the calibration to be inaccurate. Where do the errors potentially reside?


The Calibration & Verification Process

The concept of verifying a calibration by running a second profile sequence should be rather obvious, and in simple terms makes perfect sense.

After first profiling the display, following the Initial Display Set-Up, 3D LUT Calibration, and the Hardware Integration specific User Guides as needed, and generating the final 3D Calibration LUT, any verification process should then prove the accuracy of the calibration LightSpace has performed. Especially if the calibration workflow has been defined to include the entire image path, and not deal with the display in isolation. It should compensate for any signal path errors, should it not?

Unfortunately, the answer is often no...



Direct HDMI Calibration & Verification

To understand potential Calibration & Verification Errors it is best to initially look at the simplest calibration workflow, using a display with internal LUT capability and profiling via a direct HDMI connection from the LightSpace PC.

Direct Calibration

With a Direct profiling configuration the HDMI output from the LightSpace laptop is connected to the display's HDMI input, enabling direct profiling, as described in the Direct HDMI Profiling section of the Direct Profiling User Guide.

It is also assumed the HDMI signal is RGB, and NOT YCbCr - see later!

In this profiling workflow the RGB triplet patch values, as defined by LightSpace, are directly displayed on the monitor via the HDMI connection (assuming there are no active ICC profiles, or incorrect Graphics Card/Chip Set configurations, as defined in the Direct Profiling User Guide), ensuring the values measured by the probe are as expected for the TPG triplet as defined by LightSpace, and as accurate as the probe's inherent capabilities.

Based on this direct profiling, the 3D calibration LUT generated will, in turn, be as accurate as it can be, based on the probe's reading accuracy, and the stability of the display.

LUT Verification

After LUT Generation, the first verification step is to run a Quick Profile using the Active LUT capability within LightSpace.

The Active LUT capability effectively 'proves' the underlying LUT accuracy by passing the patch colour data through the LUT, before sending the patch to the display via the HDMI connection.

As a result, the patch colour displayed will have been 'corrected' by the LUT, and the probe will read the new, corrected value, effectively proving the accuracy of the calibration.

The next step is to Upload the LUT into the display (as for this example the display has 3D LUT capability), and again re-profile.

Ideally, both verification profiles will match exactly, and the calibration will be perfectly accurate.

So, what could be wrong?

In this example, probably very little, as the workflows is very simple, assuming the instructions on the management of ICC profile and Graphics Card/Chip Set configurations have been performed correctly.

Even though the Active LUT and the actual LUT Upload into the display are in different location the simple RGB signal path minimises potential issues.

The following shows the signal and profiling patch for the 'Active LUT'.


Patch Generation Patch Generator Active LUT HDMI Connection Display Measurement
Patch colour defined by LightSpace, and sent to display as a HDMI RGB signal LightSpace Test Patch Generator (TPG) located before the LUT Active LUT held with LightSpace, modifies the patch colour before sending to the display HDMI RGB connection to the display Active LUT corrected patch colour displayed and measured on monitor
LightSpace Laptop TPG Colour Space Conversion LUT HDMI Cable Working Colour Space

And the following shows the signal and profiling path when the LUT is uploaded directly into the display.


Patch Generation HDMI Connection Patch Generator Uploaded LUT Display Measurement
Patch colour defined by LightSpace, and sent to display as a HDMI RGB signal HDMI RGB connection to the display Display internal Test Patch Generator (TPG) located before the LUT LUT uploaded into the display modifies the patch colour LUT corrected patch colour displayed and measured on monitor
LightSpace Laptop HDMI Cable TPG Colour Space Conversion LUT Working Colour Space

As can be seen, there is little to go wrong, as the LUT will have the same effect on the signal as an 'Active LUT', and when 'Uploaded' into the monitor. And as the signal profiling path is direct and simple, with no potential for distortion, the final calibration LUT will be as accurate as possible.

If there are any differences between the two verifications that would point to some form of variation in the LUT when uploaded into the display vs. the Active LUT held within LightSpace. Such variations could be LUT size (say, 17^3 vs. the internal LightSpace 33^3 size), or the need to apply a VideoScale process to the LUT before uploading, if the signal path is TV Legal, and required video scale data within a data range LUT. See VideoScale for more information.

Any 'gross' errors, with the calibration appearing incorrect in any way, are likely down to active ICC profiles or VCGT (Video Card Gamma Tables), and should be simple to spot as both verifications will be wrong/poor.


Internal TPG Calibration & Verification

The first real potential for error is if the display has an internal test patch generator (TPG), as it is possible that the internal patch generator may interpret the RGB triplet patch values sent by LightSpace incorrectly. Therefore the measurement values recorded by the probe would not match the correct calibrated values, and so provide inaccurate calibration, and inaccurate verification.


Patch Generation Data Connection Patch Generator Uploaded LUT Display Measurement
Patch colour defined by LightSpace, and sent to display as RGB Triplet values via Network/USB connection USB/Network data connection to the display Internal Test Patch Generator (TPG) located before the LUT LUT uploaded into the display modifies the RGB Triplet colour data LUT corrected patch colour displayed via internal display TPG and measured
LightSpace Laptop Data Cable TPG Colour Space Conversion LUT Working Colour Space

The potential for errors can stem from the display being YCbCr based, as with some broadcast monitors, with errors in the conversion of the LightSpace RGB Triplet patch values into YCbCr values for the TPG, and also within the re-conversion of the YCbCr values back into RGB before passing through the calibration LUT (as all 3D LUTs are RGB based) and then displayed on the screen (again, as all display screens are RGB).

If the RGB to YCbCr and back to RGB colour space conversion is not accurate, the calibration and verification will not be accurate, as can be seen below using RGB Balance graphs to show Grey Scale variations.

Native Display - Internal TPG

The native display shows a large blue contamination in the blacks/shadows, with a predominantly green grey scale.

Active LUT Verification

The calibration LUT when verified as an Active LUT within LightSpace shows a reasonable Grey Scale calibration.

LUT Upload Verification

When the same LUT is Uploaded into the display, and re-verified, the results are less accurate.

Native Display - External TPG

When profiled with an external TPG the native display response is similar to the Internal TPG results, but not identical.
(Click HERE to see a direct comparison.)

Active LUT Verification

The calibration LUT when verified as an Active LUT within LightSpace again shows a similar result to the Internal TPG, but not identical.

LUT Upload Verification

And when the same LUT is Uploaded into the display, and re-verified, the results are again less accurate, but also again different to the Internal TPG results.

What the above errors show is that this monitor's signal path processing is less than ideal, with direct Internal TPG vs. External TPG differences, as well as Active LUT vs. Uploaded LUT variations.

It is also possible for the 'error' to be the other way round, when an internal TPG is used to perform the initial profiling and calibration LUT generation. This time the issue is the display's internal signal processing electronics distorting the input video signal.

In the following example the internal TPG has been used to profile the display and generate a calibration LUT. With the Active LUT held within LightSpace the calibration appears less accurate compared to when the same LUT is actually uploaded into the display.

Native Display - Internal TPG

The original display profiled with its own internal Test Patch Generator.

Active LUT Verification

The calibration LUT when verified as an Active LUT within LightSpace shows significant errors.

LUT Upload Verification

But when uploaded into the display the calibration results are greatly improved.

The causes of the inaccurate verifications are down to the display's video processing electronics distorting the internal TPG output signal, and when the LUT is held within LightSpace as an 'Active LUT' it is generating corrected patch values 'before' the patch generator, and so not correctly compensating for the TGP errors.

When the Calibration LUT is Uploaded into the display, it is located After the TGP, and so correctly compensates for the signal path errors after the internal TPG.

Note: with any display it is really preferable to use a RGB signal path to avoid potential colour space conversion issues, as well as issues associated with 422 colour sub-sampling.


TPG Location Issues

Following on from the above potential issues with internal TPG use, there are also potential issues with the location of the TPG - some very obvious, others more difficult to understand.

For example, Resolve, some Eizo displays, and other manufacturer's displays, have one very specific issue due to the location of the in-built patch generator.

The very obvious issue is that if the patch generator is located AFTER the 3D LUT, it cannot be used to verify the LUT calibration. It can only be used to profile the display per-calibration.


Patch Generation Network Connection Uploaded LUT Patch Generator Display Measurement
Patch colour defined by LightSpace, and sent to display or Resolve as RGB Triplet values via Network connection Network data connection to display or Resolve TGP LUT held within display or Resolve Test Patch Generator (TPG) located AFTER the LUT Patch colour displayed on monitor unchanged by the LUT
LightSpace Laptop Network Cable Colour Space Conversion LUT TPG Working Colour Space

The obvious way to 'verify' a LUT when the available TPG is after the LUT location is to use the 'Active LUT' capability within LightSpace. However, you must be aware of the possible accuracy issues due to unexpected image processing distortions.

Another way to overcome this issue when using a 3D LUT within Resolve is to use LightSpace's DIP (Display Independent Profiling) mode, and build a 'Patch Set' timeline to play the calibration patches in real-time, in sync with LightSpace, as any images on the timeline will pass through the LUT.

There are also other possible compounding issues, as if there are any signal 'changes/modifications' prior to the TPG, even when the display is 'un-calibrated', such signal changes will not be compensated for within the LUT, making the LUT inaccurate.


Patch Generation Network Connection Video Processing Patch Generator Display
Patch colour defined by LightSpace, and sent to display TPG as RGB Triplet values via Network connection Network connection to the internal TPG Video electronics distorts 'normal' image signal, but is bypassed by the TPG signal Internal TPG has no knowledge of the signal processing distortion Patch colour displayed and measured on monitor
LightSpace Laptop Data Cable Monitor Electronics TPG Working Colour Space

Such issues will show as variations in the LUT verification, when verified using Active LUT vs. with the LUT uploaded into the display, as defined within the section on 'Signal Path Issues'.


LUT Size Issues

Looking at the above variations between the LightSpace Active LUT and the Uploaded LUT there is one other potential area for error - the actual LUT size.

The default LUT size within LightSpace is 33^3, but when uploaded into a display with internal 3D LUT capability the size can potentially be different, and will therefore show some variations in the verification results.

33^3 LUT

The above is a standard LightSpace 33^3 LUT calibration result.

17^3 LUT

This shows the result of using a 17^3 LUT, and the main variation is more blue contamination on the blacks/shadows.

5^3 LUT

And this shows a very small 5^3 LUT, and again the main variation is even more blue contamination on the blacks/shadows.

As can be seen, the likely variation in the final calibration result is predominantly in the LUT's ability to manage the native excessive blue in the blacks/shadows of the display. There are not the same grey scale errors as seen in the Internal TPG Calibration & Verification issues above.


Signal Range Issues

The expected signal range for any given display - TV Legal vs. Data Range - can also cause calibration errors, if the correct range workflow is not applied.

Issues can be caused both when profiling, as well as when the LUT is applied.

Correct Range - Gamma
Correct Range - RGB Balance
Correct Range - Clip

The above graphs should be used as a 'reference' for the graphs that follow, not as ideal plots in their own right.

Incorrect TV Legal Range Profiling - Gamma
Incorrect TV Legal Range Profiling - Balance
Incorrect TV Legal Range Profiling - Clip

In the above graphs the display has been profiled with a TV Legal range signal, when the display expect a Data Range signal. The gamma plot has shifted up, while the RGB Balance shows less 'blue in the blacks', and the clip graph shows a lifted profile, due to the gamma shift.

The real issue is that the black level has risen, and the white dropped, resulting in a reduced contrast range overall.

Incorrect Data Range Profiling - Gamma
Incorrect Data Range Profiling - Balance
Incorrect Data Range Profiling - Clip

In the above graphs the display has been profiled with a Data range signal, when the display expect a TV Legal signal. The gamma plot has shifted down, while the RGB Balance shows a step in the blacks, and the clip graph shows an obvious clipping error, due to the two lower grey scale patched being identical.

Obviously, any calibration LUT generated with incorrect range profiles will cause incorrect calibration.

Additionally, even if the LUT is generated with the correct range profile, issues be caused when uploaded into the display if the LUT is not scaled to match the signal path. The most obvious issue is when the 3D LUT within the display (or LUT Box!) uses a full-range cube, but the signal path is TV Legal, meaning that the LUT data needs to be scaled to match the TV Legal range within a data range LUT, using VideoScale. See VideoScale for more information.

Incorrect LUT Scaling - CIE
Incorrect LUT Scaling - Gamma
Incorrect LUT Scaling - Balance

The above graphs show the results of an incorrect scaled LUT, with the LUT being Data Range for a TV Legal workflow.

Correct LUT Scaling - CIE
Correct LUT Scaling - Gamma
Correct LUT scaling - Balance

The above show the correct results when the LUT has been scaled correctly using VideoScale.


Signal Delay Issues

A very obvious, but easy to overlook issue is that of a signal delay from the LightSpace PC to the actual display being calibrated, especially when the connection is not a simple direct HDMI connection.

Such a delay means the the patch being displayed on the monitor being calibrated actually occurs slightly after the patch displayed within LightSpace. Therefore, unless the correct amount of Extra Delay is added, the probe will actually start reading before the correct patch is being displayed.

Extra Delay

Any calibration or verification result is therefore going to be wildly incorrect, and should be simple to spot, as most of the graphs will show wildly inaccurate plots.

Built into LightSpace is an 'Extra Delay' option, to add the necessary delay into the initiation of each probe measurement to counter the signal path delay.


Signal Path Issues

Signal path issues are where the display's internal video processing 'distorts' the video signal in an unexpected, and inaccurate way, and can cause even great problems if the image path is more complex, and uses a number of different hardware components.

For example, the following shows the RGB Balance verification from a display connected to a Resolve system, using a BMD Decklink card, compared to the same display calibrated using a direct HDMI connection.

Native Display

The native display shows a large blue contamination in the blacks/shadows, with a predominantly warm (red/magenta) grey scale.

Resolve & Decklink Calibration

When profiled, calibrated, and verified via a Resolve system, using a Decklink card, the results are not as expected, with an obvious green/blue error in the shadows.

Direct HDMI Calibration

When the same display is calibrated and verified via a direct HDMI connection to the LightSpace PC the results improve, with a near perfect grey scale.

In the above case the issue is incorrect RGB to YCbCr conversion in the signal path, with the RGB Triplet values from LightSpace being converted to YCbCr within Resolve, and then within the display the YCbCr signal being re-converted back to RGB.

Somewhere within those colour space conversions there is an error, that while not immediately visible under normal display use, becomes very obvious during attempted calibration.

The solution is to set Resolve to RGB444, both with the Resolve system, and with the Decklink card, and set the display to explicitly accept RGB input.

Note: The conversion error may be within Resolve, the Decklink card, the display, or all three... Without using an external TPG, such as the Murideo SIX-G that can be set to accurate YCbCr and RGB patch generation, it is impossible to define where the issue resides.

It is worth assessing Signal Path Issues in greater detail, as they can be the root cause of a lot of inaccurate final calibrations, so understanding in more detail the workflow of potential problems will help with understanding.


Pre-Calibration Profiling
Using direct HDMI from LightSpace to profile the display to enable the generation of a calibration LUT.
Patch Generation Patch Generator HDMI Connection Video Processing Display
Patch colour defined by LightSpace LightSpace Test Patch Generator (TPG) sends patch via HDMI to display HDMI RGB connection to the display Video electronics distorts patch colour in an unexpected/inaccurate way Patch colour displayed and measured on monitor
LightSpace Laptop TPG HDMI Cable Monitor Electronics Working Colour Space

In the above pre-calibration workflow the video processing electronics adds a distortion to the patch colour, before it reaches the displays screen (the actual gamut and gamma of the display is irrelevant - it is just the additional inaccurate processing distortion that is important - and that can be within the display, or any external video processing system within the signal path.

As a workflow example, a skin tone patch with triplet value (145, 107, 113) is defined within LightSpace, and sent to the display. The video electronics (which could be an external patch generator, LUT box set to Bypass/null, or within the actual display) distorts the triplet values to 146, 105, 114, and it is that triplet value that is then displayed on the screen and is the resulting colour measured by the probe.

When LightSpace generates a calibration LUT it will generate a 'correction' within the LUT that corrects for the colour the display showed, which includes a correction for the distortion introduced by the video processing electronics.

However, the LUT correction will have a different result depending on where in the video path the LUT is then located...

If the LUT is positioned before the video processing electronics it will apply the correction to the generated colour in advance of the error being introduced. So, the 145, 107, 113 value will be come 144, 109, 112, and when this colour reaches the video processing electronics, the error introduced will most likely not be the same 'relative' error as before, as the colour triplet value is NOT the same! So the LUT correction will now be wrong, as the introduced error will be different!

The LUT must be positioned AFTER the video processing electronics to correct the introduced errors.


Gamut Coverage

Measuring gamut coverage, using just the peak chroma values, is not a good way to measure calibration accuracy, as greater peak chroma gamut coverage does not always equate to better accuracy.

Native Display Gamut

The native display gamut suggests it can cover a fairly large percentage of the target Rec709 colour gamut.

Greater Gamut Coverage?

This calibration appears to show that a greater gamut coverage is possible, but note the peak green measurement is not correct, as it is 'off-axis' with respect to the target colour space. The Green Hue is inaccurate.

Greater Calibration accuracy!

When calibrated with LightSpace Peak Chroma the gamut coverage is lower, but the peak green measurement is accurate with respect to the target colour space. The Hue is now correct.

Calibration Error

When compared directly the calibration inaccuracy is obvious.
(We are focusing on the Green error, although the Blue error is also plotted.)

The assumption that is all to easy to make is that the area outside the peak primary gamut triangle is going to be 'cut off', and therefore drastically reduce the display's colour range.

This is not correct, as the gamut triangle is ONLY showing the values for the peak chroma values, not the gamut coverage for any other colours, and LightSpace will calibrate each potential colour separately, maximising the total volumetric calibration of the display.

Gamut Clip?

The immediate concern is that all colours within the shaded region will be cut-off, reducing the display's overall gamut coverage and reduce the displayed colour range.

Beyond Peak Chroma Gamut

However, all colours that the display can accurately show will be calibrated correctly, as can be seen by measuring a 100% green patch, with the addition of 75% red, so placing it on the edge of the target Rec709 colour space, which is outside the peak colour gamut triangle.

Volumetric Gamut

This can be further proven by using a full volumetric (cube based) profile, to show the full volumetric gamut coverage, which is not constrained by the peak chroma gamut.

As can be seen, the Peak Chroma gamut triangle is a poor representation of the actual volumetric gamut accuracy of the display, and should not be relied on as a guide for the actual colour range of any calibrated display.


Matrix or 3D LUT

The above Gamut Coverage information brings us to another often asked question, that while not really an 'error' does impact the accuracy of calibration.

That question is 'what is the difference between a 3x3 Matrix (with or without a 1D LUT), compared to a 3D LUT calibration?'

The above is a very good example of the difference, as a matrix calibration will not be able calibrate all available colours the display can accurately display, while a 3D LUT calibration will. A matrix calibration will clip the display gamut to the peak colours.

3D LUT Calibration
3D LUT Gamut

Matrix Calibration
Matrix Calibration

The 'clipping' is due to a 3x3 matrix defining a six sided 3 dimensional shape, with flat sides, while a 3D LUT will map each and every input colour to a separate output colour, without constraint, meaning each individual 'colour' can be 'calibrated' without limitation or being restricted to peak primary colours (as the above Gamut Coverage section shows).

The following 3D Cubes show the difference between the 3D LUT and 3x3 Matrix, with the Matrix unable to allow calibration outside of the calibrated primary colours.

3D LUT based Cube
3D LUT Cube

Matrix based Cube
Matrix Cube


Quick Profile Calibration not accurate

Using a Quick Profile (Grey only, or Primary Only Quick Profiles are all that is needed) to perform a 3D LUT based calibration can be a very quick, and potentially very accurate calibration method.

However, Quick Profile based calibration will only work accurately if the display has a very linear response to input signal changes, as well as good RGB Separation, and potentially good RGB Balance.

The problems arise with understanding just what the above statement really means...

The first issue is display response linearity to input signal change, which means the display on-screen output changes by an amount that is equal to any input signal change. For example a change in an input colour triplet value to a brighter value, with no change to the actual colour, causes an equal brightness only change on-screen, with no change in the colour.

Issues with display response linearity can usually be seen with a Primary & secondary Quick Profile, as follows.

The following left-hand graph shows an example of a display with a very non-linear response to input signal changes. Using a Primary and Secondary Quick Profile it can be seen the for red, Yellow, Cyan and Magenta the hue of the measured patches tracks along the edge of the gamut triangle, even though the hue for each input colour patches is identical.

Non-Linear response
Quick Profiling

Linear Desaturation
Quick Profiling

The right-hand graph shows a response that is a linear desaturation of of the output colour as luma decreases. The measured patches maintain the same relative hue, and track towards the native backlight colour temperature of the display. Such a response will not prevent the use of a Quick Profile for LUT generation.

RGB Separation compares each primary R, G, and B patch of the same stimulus value (for example Red 128,0,0, Green 0,128,0, and Blue 0,0,128) to the equivalent grey scale patch (128,128,128), matching the individual RGB patch measured values to the expected colour matrix combination for the equivalent grey patch. Any error in the graph shows the display is suffering colour decoupling issues with the display's separate RGB colour channels. This means that an input colour change that should affect only a single colour channel also causes changes within the other colour channels - what is known as cross-coupling between the colour channels.

Reasonable RGB Separation
Quick Profiling

Poor RGB Separation
Quick Profiling

Channel cross-coupling is an extreme form of display non-linear response, and any display suffering this will require a full volumetric Cube based profile for accurate calibration.

While the left-hand graph shows 'reasonable' RGB Separation, in reality for the best calibration results even a display with this response should be calibrated with a cube based profile.


Display Technologies

An obvious issue with different display technologies is Metameric Failure, as described within the Perceptual Colour Matching page.

However, there are other issues that are less obvious, but can cause image problems that can be very difficult to overcome.

As part of the assessment of display calibration we have adopted a 'self-profile' approach to assessing any display, to better understand a given display's underlying capabilities and issues that will affect potential calibration accuracy.

This concept stems from a very simple premise - any 'good' display, when in its native, uncalibrated state, should effectively profile to itself very accurately.

Therefore, measuring the RGB primaries, and white point, for any display, and generating a target colour space with those values, a full volumetric profile of the display should map each and every measurement accurately to the target colour space.

With LightSpace this is relatively easy to perform, by profiling the display with a large cube based profile (with the display set to its native, uncalibrated setting), generating a new Colour Space with the peak RGB & W values, as well as gamma, taken from the profile, and then generating a LUT with the Source as the new colour space, and Destination as the actual profile. The closer the LUT is to 'unity' the better the underlying capabilities of the display.

The following 3D Cubes show the difference between the 3D LUT generated for a WRGB OLED, and RGB LED Backlight LCD.

WRGB OLED
WRGB OLED Display

RGB LED Backlight LCD
LCD Dispaly

As can be seen, the WRGB OLED suffers from non-linear (random) colour inconsistencies, while the LCD display has a more linear/uniformed volumetric colour profile throughout.

Looking at the CIE graphs, and comparing the Primary RGB points with the Secondary CMY points cal also shows non-linear inconsistencies, as on good displays the Secondary CMY points are always a direct calculation of the Primary RGB points, so will track in an identical way.

In the following graphs the Yellow and Magenta Secondaries of the WRGB OLED are tracking along the edge of the CIE gamut triangle, while the RGB Primaries, and the Cyan Secondary, show no such 'tracking' issues.

The LCD display show no such issues, just showing the expected saturation tracking of both Primary and Secondary colours with increasing brightness (OLEDs maintain their saturation throughout their brightness range, which can make the image feel 'unnatural' - see: Advanced LightSpace CMS Operation, the section on OLED Dislays.)

WRGB OLED
WRGB OLED Display

RGB LED Backlight LCD
LCD Dispaly

Such issues can be further seen by looking at the Delta-E distribution graph, again with the profile referenced to the actual native colour space of each display. The colour inconsistencies with the WRGB display causes a greater spread of dE values, while the more inherently linear (accurate) LCD display has a better concentration of low dE values.
(The more to the 'left' and 'high' the dE graph, the better the display accuracy.)

WRGB OLED
WRGB OLED Display

RGB LED Backlight LCD
LCD Dispaly

Profiling a display to itself (to its own native Gamut/Gamma) really does show the underlying quality of any display!

And a final verification of the inherent instabilities can be seen with 'Drift' graphs and a Sequential pathc sequence, which show thermal drift with WRGB OLEDs, combined with a colour temperature change over time as shown by the change over time of the RGB plots, with the display starting warm (red/yellow), and then ending cold (blue/cyan). Again, the LCD display shows a far greater level of stability, and negligible colour temperature drift over time.

While the thermal drift in the WOLED can be improved by using a different patch set (LightSpace Anisometric vs. Sequential, Drift and Stabilisation patch insertiopn), the fact the display shows issues with a Sequential sequence show the underlying issues.

WRGB OLED
WRGB OLED Display

RGB LED Backlight LCD
LCD Dispaly

For displays with inherent underlying non-linear inaccuracy, as shown above when a display is profiled to itself (and not directly associated with thermal instability, which requires the use of 'Stabilisation' patches to overcome), a very large profiling sequence will be required, in combination with a very large 3D calibration LUT to overcome the non-linear issues.

WRGB OLED technology will inherently suffer such issues, due to the inclusion of the 'white' pixel, as this will distort the standard RGB colour channel relationship.

Any display that shows any of the above stability issues really should be treated with caution if attempted to be used in colour critical environments, such as for grading work.

For additional information on WOLED issues see: The Light Illusion Forums.


Post Calibration Accuracy

A question that is rarely asked, but in reality is critical for any display, is how well calibrated is the display AFTER calibration..?

Obviously, any display will drift in time, and so will require re-calibration, but that is not the question here. The question is: immediately after calibration, and after the calibration has been verified to be accurate, is the display now REALLY accurate when displaying real images/footage?

You may think this a stupid question, as the display has just been calibrated, and verified to be accurate...

But, when profiling and calibrating did you use any of the 'compensation' options within LightSpace? For example, did you use Drift, or more specifically the 'Stabilisation' options?

If the answer is Yes, you must now realise that the display was only 'accurate' because of the use of those options, and when playing real footage will not have the same 'corrections' applied, and will immediately become uncalibrated.

This is especially true of displays that require the use of the 'Stabilisation' option (although also true with Drift, to a lesser extent), as the insertion of black frames after each real calibration patch allows the display to 'cool', preventing thermal instability.

Obviously, when in real use, such displays will not have regular 'cooling' periods, and will immediately suffer thermal instability, especially if the display is being used in a colour critical environment, such as for grading, as the same image/scene will be held on the display for extended periods, causing thermal instability as the display screen heats up.

This means displays that require the use of Stabilisation Patches for calibration should be treated with caution if being used in colour critical grading environments.


Post Calibration Adjustments

After performing a 3D LUT based calibration users will often be tempted to 'tweak' the display's manual controls. Such adjustments will always 'break' the 3D LUT calibration, regardless of where in the display's image processing chain the controls exits.

There is an incorrect assumption that having the display's manual controls 'after' the 3D LUT means they can be used post calibration, without affecting the accuracy of the 3D LUT calibration. That assumption is incorrect, as while not as bad as having the manual controls before the 3D LUT, altering controls post LUT will still break the calibration result.

The issue is that the 3D LUT is 'correcting' any non-linear irregularities in both the display's image processing electronics, as well as the actual screen. Therefore making any changes either before or after the LUT will break the relationship between the correction the 3D LUT is generating, and the location of the errors within the display.


Manual Controls pre-LUT
The most obvious issue is with manual controls pre the 3D LUT.
Input Signal Manual Controls Input 1D LUT 3D LUT Screen 1D LUT Screen
Input video signal Display manual controls Input 3x 1D LUT for gamma and colour temp 3D LUT form gamut calibration 1D LUT for screen linearity Display screen
HDMI Cable Manual Controls Input LUT Colour Space Conversion LUT Screen 1D LUT Screen 1D LUT
The input 3x 1D LUT and 3D LUT are effectively a single component
(often just a 3D LUT is used)
The 1D LUT can be considered part of the screen, effectively managing screen linearity
(Not all displays have a 1D Screen LUT)

In the above signal path the issues are relatively obvious, as this is exactly as per the Signal Path issues, described above. Any change to the input signal the 3D LUT expects will immediately break the calibration.


Manual Controls post-LUT
The less obvious issue is with manual controls post the 3D LUT.
Input Signal Input 1D LUT 3D LUT Manual Controls Screen 1D LUT Screen
Input video signal Input 3x 1D LUT for gamma and colour temp 3D LUT form gamut calibration Display manual controls 1D LUT for screen linearity Display screen
HDMI Cable Input LUT Colour Space Conversion LUT Manual Controls Screen 1D LUT Screen 1D LUT
The input 3x 1D LUT and 3D LUT are effectively a single component
(often just a 3D LUT is used)
The 1D LUT can be considered part of the screen, effectively managing screen linearity
(Not all displays have a 1D Screen LUT)

In the above signal path, the issues are less obvious as it is often assumed the screen has a purely linear response to input signal changes. But, that is not the case, as is defined by the need to have an input 1D LUT to attempt to linearise the screen. As a result, changing the output from the 3D LUT via manual controls (brightness, contract, gain, etc) before the signal gets to the screen will break the final display calibration in the same way as having the manual controls before the 3D LUT, although potentially with lesser obvious visual issues.

In very simple terms, making any adjustments to a display post-calibration will break the calibration.


BT.1886 Gamma

BT1886 can be a problem on displays with lifted blacks, due to the way it raises gamma in the shadow region. This causes the shadows to become washed-out, which in turn will cause the colourist to attempt to grade the shadows darker. When the graded footage is then viewed on a display with a standard power law gamma, or a display with a lower black and calibrated to BT1886, the shadows will appear crushed.

This can be seen in the following images, showing the target gamma for a display with a 0.05 nit black level, and 100 nits peak white. The lift in the shadows can easily be seen (click on any of the images, and then cycle through them).

2.4 Gamma
2.2 Gamma
BT.1886 Gamma

This can be seen in more detail in the following image, which shows the BT.1886 Gamma (blue plot) as a differential plot compared to a 2.4 power law gamma (black plot). Each plotted BT.1886 point also has its (approx.) gamma value defined.
(Note: The first and last points have no valid gamma value.)

BT.1886 Gamma

As can be seen, as the curve heads towards black the gamma value decreases, lifting the shadows.
(As the graph is a differential graph, it shows the variation from the target gamma.)


HDR Volumetric Accuracy

The accuracy of HDR displays is something of a difficult area for many to assess, as the existing metrics used to define calibration accuracy are just not capable of showing true volumetric issues.

To explain this, see the following examples generated with special display assessment software used in-house by Light Illusion. Both displays are HDR, with a Peak Luma of around 700 nits to 800 nits.

In these charts we have adopted the 'self-profile' approach defined previously, enabling a better understanding of a given display's underlying capabilities and issues that will affect potential calibration accuracy. Self Profiling also enables any display to be directly compared to any other display as a direct relative comparison.

As described before, this concept stems from a very simple premise - any 'good' display, when in its native, uncalibrated state, should effectively profile to itself very accurately.

Therefore, measuring the RGB primaries, and white point, for any display, and generating a target colour space with those values, a full volumetric profile of the display should map each and every measurement accurately to the target colour space.

The 3D charts below are also colour coded, with green measurements showing points that have a sub-1 dE. Orange points are between 1 and 2.3 dE. Red points are above 2.3 dE.

LCD HDR Display
WOLED HDR Display

In the above 3D CIE graphs the LCD display shows a relatively good/acceptable level of underlying display volumetric capability, while the WOLED graph shows issues throughout, that become increasingly worse as display brightness increases.

LCD HDR Display
WOLED HDR Display

The second set of graphs have dE tangent lines included, which show the dE error for each and every point. The difference is obvious.

LCD HDR Display
WOLED HDR Display

The 'Cube graphs are 'normalised' versions of the CIE graphs, and help visualise the volumetric issue with any display. The WOLED graph shows just how much volumetric accuracy is missing.

LCD HDR Display
WOLED HDR Display

And again, we can add dE tangent lines to help highlight the errors.

LCD HDR Display
WOLED HDR Display

The last graphs are also available in LightSpace, and really simplify displays the issue. As brightness increases the WOLED RGB gamut can't maintain an equal level of brightness with the grey scale (Luma), and clips harshly well before the peak luma of the display.

As can be seen, the LCD is 'ok', while not perfect, but the WOLED has major issues that are easily apparent, due to the inclusion of the 'white' pixel, as this distorts the standard RGB colour channel relationship - excessively at HDR brightness levels.

Standard RGB OLED displays work well for SDR use, requiring no additional white pixel. Adding the white pixel cause volumetric issues throughout the whole brightness range (including SDR), and causes serious issues with HDR. While the SDR issues can be calibrated out with a high-density 3D LUT via an external LUT box, the HDR issues cannot.
(To see SDR volumetric issues issues see: The Light Illusion Forums.)

In very simple terms, this means is the WOLED can never be calibrated for HDR...


Additional Technical & Support Info.

German language version of this page: displaycalibration.de

© Light Illusion - All rights reserved - Privacy Notice