| Forums | Register | Reply | Polls | Search | Statistics |
Display Calibration Light Illusion Forums / Display Calibration /  

Panasonic FZW804 OLED HDR Calibration

Author holger
#1 | Posted: 15 Sep 2018 14:09 
Recently i have bought a Panasonic OLED TV (2018 model). I have calibrated it for SDR using lightspace, everything good so far. Now, after i have read a bit about HDR Calibration, i should rather write HDR "Calibration" as it seems that the result heavily depends on "defeating" the "tone mapping" that is going in the TV but still get the panel into "HDR Mode" (Maximum possible luminance). I am also the owner of a radiance 2143 at the moment but want to upgrade to a pro (the plan is to use its "intensity mapping" feature together with a lightspace generated 3D LUT). Today i tried a HDR capable movie player with the TV to evaluate what "controls / settings" there are in HDR mode. Now there is one setting that looks interesting - "HDMI EOTF TYPE" which is configurable per HDMI input. The options are "PQ / HLG / Traditional Gamma" and i figured that maybe setting this to "Traditional Gamma" does what i want / need to get a usable HDR calibration.

Any thoughts / ideas / corrections are welcome here .. thanks

Author gordonF
#2 | Posted: 16 Sep 2018 10:03 
I've not tried this with the Panasonics but using the HDR flag option in the Pro CMS section should be able to do what you want. IE Send a 3DLUT with expectation of gamma 2.4 to the display but also send HDR flag to make the panel go to HDR mode. Then set the TV to normal gamma 2.4 and away you go. I asked for this feature specifically for use with JVC projectors as they have similar capability.

Author Steve

#3 | Posted: 16 Sep 2018 12:06 
I've not seen the Panasonic FZW804 either, but if it can be set to 'Traditional Gamma', and maintain its peak (HDR) luma output, without the need for HDR metadata in the video signal, you can then use a HDR LUT in any LUT box...

However, 'Traditional Gamma' may disable the TV's HDR mode.

Steve Shaw
Mob Boss at Light Illusion

Author ConnecTED
#4 | Posted: 17 Sep 2018 09:47 | Edited by: ConnecTED 
There problems with HDR and 3D LUT in consumer world.

These problems are external 3D LUT Box agnostic, it doesn't matter if your TV has internal 3D LUT capability for HDR, like Panasonic EZ1000 for example, which has internally 3D LUT and 1D LUT tables for SDR and for HDR modes, with any method its impossible to calibrate for 3D LUT and HDR10 in consumer display world.

There many reason why is not possible, the main reason for OLED is due to WRGB pixel structure.

WRGB OLEDs due to the introduction of the 'white' sub-pixel, this distorts the standard RGB color channel relationship - excessively at HDR brightness levels. (if you sum the Y of 100% patch of R+G+B primaries you get 400nits while the same time if you display a 100% White patch you get 800nit...so your color gamut is limited to 400 nits... this means that WOLEDs can never be calibrated for HDR... ...but can be calibrated with 3D LUT in SDR mode, the recommendation is up 105-110 nits, there will be to ABL limiting and displays are more stable overs the time at these nits levels.

WRGB OLED's displays are NOT accurate volumetrically at all. It will require more advanced methodology from typical/limited patch-set and classic verification methods.

A great approach to assessing any display, to better understand a given display's underlying capabilities and issues that will affect potential calibration accuracy is to profile the native (un-calibrated display state) to itself, so measuring Primary colors and white point and generating a target color space with those values, a full volumetric profile of the display should map each and every measurement accurately to the target color space.

For LightSpace users, its relatively easy to perform, by profiling the display with a large cube based profile (with the display set to its native, un-calibrated setting), generating a new Color 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 color space, and Destination as the actual profile.

The closer the LUT is to 'unity' (full cube with equaled spaced dots, when you will look the 3D Cube Graphics LUT Viewer) the better the underlying capabilities of the display.

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 color channel relationship.

If you look at this image...

A Piccy

...you can see the way the 'vectors' change direction, and how multiple 'input' points all point towards the same actual color point.

(The colored dot is the measured color point, color coded for dE accuracy, while the 'vector line' shows where the point should actually be...

A Piccy

The output of the display is not linear and predictable when the input signal is changed in a linear and predictable way.

The causes display volumetric non-linearity.

In its simplest form, a displayed color is simply the sum of its components C= R+G+B.

If it is 'linear', then the combination is predictable in that changes in the input red signal has no effect on the light coming from the blue or green channel.

This is mostly true for old school CRT's, and for the most part LCD's and even LED's.

With a WOLED display, it has 4 emitters.

So, when the red signal is adjust, it will change the emission from the red LED, and may be also the white LED depending on the overall RGB signal that it is receiving.

This breaks the simple C= R+G+B rule.

The device now has a complex signal to light behavior.

This means the display is very non-linear in its response.

If you have a non-linear system, you need a full 3d model to make it do what you want it to. The greater the non-linearity, the bigger the size of the 3D LUT required to manage it.

(Actually, if you had direct access to RGB and W, then you would be best generating a 4 channel cube! This is often done internally in printers that have more than 3 inks; same basic idea.)

All above data coming from SDR profiling data, stuff are lot of worse in HDR mode, lets see about this below....

... by comparing WRGB OLED @ HDR mode (without Tone-Mapping active) 800nits peak vs. LCD @ HDR mode (without Tone-Mapping active) comparison will be used to better demonstrate the problem.

To be able to perform 21-Point Cube profiling in HDR mode (without tone-mapping activated), Maciej Koper (mkoper - hdtvpolska.com); popular and very expert calibrator/TV reviewer from Poland, performed a lot of measurements using LightSpace and testing the Stabilization feature, to improve panel stability over the time.

The following shows a volumetric comparison between WOLED and LCD displays when used for HDR, with a peak Luma of 700-800 nits, again with both displays profiled to themselves to provide relative data that is easy to compare.

These examples have been generated with special display assessment software used in-house by Light Illusion

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.

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

The 3D charts below are also color 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.

As can very easily be seen the WOLED has major volumetric issues when used for HDR, as the color gamut luma cannot match the grey scale luma, causing a clipping of RGB Separation.

The LCD display has relatively consistent volumetric accuracy throughout.

A Piccy

A Piccy

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.

A Piccy

A Piccy

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

A Piccy

A Piccy

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

A Piccy

A Piccy

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

RGB Separation Charts

A Piccy

A Piccy

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 color matrix combination for the equivalent grey patch.

Any error in the graph can show the display is suffering color decoupling issues with the display's separate RGB color channels, or can show that the color channels just can't match the grey scale luma levels, as in this case.

Author ConnecTED
#5 | Posted: 17 Sep 2018 10:02 
Again, looking at the above, a good display would profile to itself near perfectly... and would have an RGB Separation graph where the RGB plots follow perfectly the target line (all overlapping).

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 color channel relationship - excessively at HDR brightness levels. (if you sum the Y of 100% patch of R+G+B primaries you get 400nits while the same time if you display a 100% White patch you get 800nit...with WOLED's)

In very simple terms, what this means is the WOLED can never be calibrated for HDR... ...but calibrated with 3D LUT in SDR mode, the recommendation is up 105-110 nits, there will be to ABL limiting and displays are more stable overs the time.

Another problem is the stability, this is not a problem for WRGB OLED's only but from consumer LCD-LED HDR displays also...they are very unstable to their HDR mode, so this makes any size of cube profiling impossible.

Think that Panasonic EZ1000 which has 3D LUT for HDR10, and there is a way to disable any internal gamut/tone mapping (while you will have gamma 2.4 750 nits...it doesn't exist of the display in consumer world where you can disable internal PQ tracking), due to large drifting and WRGB structure also.

Baked Tone/Gamut mapping can't be disabled... If you want to perform a 3D LUT for HDR10, the display need to have gamma based tracking (not PQ), to be able to provide you the max output possible (700-800 nits) with gamma 2.2-2.4, and don't have enabled any kind of tone/gamut mapping.

The generated 3D LUT table will have target PQ and REC.2020 ColorSpace and the type of tone mapping you will want (hard clip, soft clip, when the roll-off with start etc...) all these details its something the software will calculate, LightSpace has all these tools at least 2-3 years for 3D LUT for HDR:

A Piccy

So the display will have a fixed output, it will not care about any HDR10 Static Metadata Info of each movie.

Look what is happening to HDR mode using 2% patterns (trying to not trigger ABL and load) this is the LightSpace's drift plot of White patch measured every 30 seconds (axis are +-3 nits), when Klein K-10A used for measurements of 17-Point Cube for HDR Mode with Peak output 796 nits:

A Piccy

The best possible scenario, which has been tested, is to use 9 seconds of Black frame insertion before each measured patch, so it took 26H to complete the 21-Point Cube profiling, in a very cold room, with fan aiming the panel also to cool it down... (while it takes about 2H normally with K-10A without black frame insertion for 21-Point Cube measurements with Klein K-10A and LightSpace).

So with 9 seconds of black frame before each meter read, the Drift Plot improved as you can see:

A Piccy

...but is still unstable, and it was impossible to get a valid 3D LUT for HDR because the panel's unstable performance and the issue with W sub-pixel which is reducing a lot the gamut color volume.

This is a proof that 3D LUT with WRGB (with disabled tone mapping and with display which has 3D LUT for HDR capability) is not working.

Maciej Koper (mkoper - hdtvpolska.com), popular and very expert calibrator/TV reviewer from Poland, performed a lot of testing for dates to find out the best number for stabilization time and helped for these data.

Since it's impossible to calibrate 3D LUT for HDR10, we will wait to see how future quantum dot (or printable) RGB OLED's will perform about stability in the future (to prevent at least the 1st issue, with W sub-pixel), and if they give the capability the companies to disable the internal tone/gamut mapping, we will test again in the future.

Author ConnecTED
#6 | Posted: 17 Sep 2018 10:05 
So it really don't matter for consumer displays any of the tricks the UHD players (like Panasonic) are doing in consumer HDR displays, they make sense for projector users only.

With consumer displays to get high output levels, this is possible when the display will receive HDR10 metadata, unless you send HDR10 metadata, you can't get high output, you will be limited to 350-400 nits max.

I have tried to create 3D LUT @ 2016 at these luminance levels with LG OLED and it was impossible. (without display entering to HDR mode)

Author holger
#7 | Posted: 17 Sep 2018 16:32 
Thanks all for your replies, ConnecTEDs reply is not very reassuring ... but it makes sense. So in essence the only real option is to go with SDR (as you wrote, 350-400 nits max and thereby avoiding the "color clipping" due to the separation of R,G,B and W) and use that for HDR as well. I wonder though what the displays internal engine does to "simulate" (color) HDR then (the technical terms are not correct here but i think you understand what i mean). Lets assume you have a very bright (>400 nits) yellow (or red or green or whatever, colourful) highlight in HDR and the display wants to show that. You are basically saying that it can't. Does it do some sort of "dithering" then and what is the perceived color and luminance then and who came up with the algorithm for that and how is it specified (well i guess it is not but proprietary "tone mapping" by Panasonic / LG for their OLED panels)? Question over questions (and a disillusioned OLED owner ...)

Author ConnecTED
#8 | Posted: 17 Sep 2018 16:48 
If you are very familiar with HDR movie playback from a perfectly manually HDR10 calibration with WRGB OLED you will see that randomly the colors are good, even if the charts are fine, when a scene will have colors that the WRGB OLED can't display you will experience a very un-natural image.

There no need also to aim for SDR and 350-400 nits, panel is super unstable at these nits.

How its display is controlling the WRGB sup-pixels is very different, there no logic where all displays are following....for example to Sony AF8, when you will display 100% White 100nits, all the RGB subpixels will be off and only the white and slightly the red will be visible. With LG, all will be open excect green to some models, all open except blue to others. There no logic, for that reason is complex to be calibrated also and require some thousand color patches to have better results.

Display Calibration Light Illusion Forums / Display Calibration /
 Panasonic FZW804 OLED HDR Calibration

Your Reply Click this icon to move up to the quoted message

Watch this topic / Get updates by Email
Only registered users are allowed to post here. Please, enter your username/password details when posting a message, or register first.


Online now: Guests - 1
Members - 0
Max. ever online: 264 [6 Jun 2020 12:02]
Guests - 264 / Members - 0