Spears & Munsil 4K UHD HDR Benchmark (Test Discs) May 15, 2023

[Stacey Spears]

Originally Posted by mrtickleuk

Thank you @Stacey Spears! I've thought of two, actually:

1. A "coming out of black" torture test for OLEDs. The sort of pattern that would trigger the "near-black chroma overshoot" problem that LG 2018 and 2019 OLEDs have which LG has worked around largely, but not eliminated completely, with firmware changes.
Similar patterns which do this are shown in Vincent's pattern here (hard to see with his pattern)

[Show spoiler]
https://www.youtube.com/watch?v=KjOb...youtu.be&t=280

and jk82's patterns here:

[Show spoiler]
https://youtu.be/Sl_t3UQ3gAk?t=73

and

[Show spoiler]
https://youtu.be/t5ZePp5_rh8

2. A screen uniformity test designed to expose "banding" on OLEDs, a bit like the one here

[Show spoiler]
https://youtu.be/dP4vEGREGD4?t=89

...but without the ridiculous "Fat Harry White" voiceover, and much much slower. Instead of rocketing through the low %age values with no time to properly see whether there's handing, I'd say about 30 seconds on each different percentage grey.

Thanks

edit: ps. I've realised I included a reference people may not get "Fat Harry White" was a parody of Barry White by a much-loved radio double-act here in the UK.
Example here:

[Show spoiler]
https://youtu.be/UpFNxhCAbE0
https://www.youtube.com/watch?v=6j7WMXJpW7E

Great feedback, thank you!

We spent a bunch of time trying to engineer a pattern for #1. Things that should not blink did and visa versa. None if them worked like we had hoped.

Now, we do have a 2D Quant pattern based on this one, which you have posted about on another forum. We made it ramp up and down from 64-90 (on the left side). It does show the flashing if you don't have the FW update installed on the LG. It also shows motion when there is none, which is caused by dynamic tone mapping. e.g. when it stops at 90, you still see movement if dynamic tone mapping is enabled. Turn it off and it goes away. Same on the bottom end, but the pause needs to be longer so the HDR10 logo goes away and does not illuminate the screen. The veiling glare makes it difficult to see. They pause for 4-seconds at each end. I think I will increase it to 6-seconds.

Here is the link.

I had not seen JK82s second pattern before. Does that show the flashing? And does it show flashing more when panning, rotating or zooming vs. just flashing?

For the other, we now have full field patterns from 0-100% in 5% steps + 1-5% in 1% steps. These will show the OLED streaking.

One of the patterns I am really excited about is pixel aging. This will allow new OLED TVs to break-in before initial calibration. You want to run it for 100 hours. It may also help with image retention.

[mrtickleuk]

Originally Posted by Stacey Spears

Great feedback, thank you!

You're welcome!

We spent a bunch of time trying to engineer a pattern for #1. Things that should not blink did and visa versa. None if them worked like we had hoped.

Pity, but many thanks for looking at it

Now, we do have a 2D Quant pattern based on this one, which you have posted about on another forum. We made it ramp up and down from 64-90 (on the left side). It does show the flashing if you don't have the FW update installed on the LG. It also shows motion when there is none, which is caused by dynamic tone mapping. e.g. when it stops at 90, you still see movement if dynamic tone mapping is enabled. Turn it off and it goes away. Same on the bottom end, but the pause needs to be longer so the HDR10 logo goes away and does not illuminate the screen. The veiling glare makes it difficult to see. They pause for 4-seconds at each end. I think I will increase it to 6-seconds.

Here is the link.

Great thanks. I will try that at home over the weekend.

I had not seen JK82s second pattern before. Does that show the flashing? And does it show flashing more when panning, rotating or zooming vs. just flashing?

It's a lot more when rotating and zooming. I think it's because it forces the TV to use different pixels for the black/not-black transitions as the pattern moves. That is also combined with the TV's temporal dithering patterns.

There is a lot less flashing with LG's workaround/fix firmware than the original firmware, though.

For the other, we now have full field patterns from 0-100% in 5% steps + 1-5% in 1% steps. These will show the OLED streaking.

Thanks, but to clarify, is this with a moving band as in that guy's video? It's the sweeping diagonal movement that really helps to show it. Much like many of the "watch scene X in episode Y" test show up the banding because the camera is moving across a statically-dark patch of scenery, and it's as that near-black scenery moves across the TV do you notice the banding.

You really don't get the same effect at all with static fields: those are good too of course, but they are not even in the same league of torture as the moving patterns are IMHO.

Example: if you have Netflix, "The OA", Season 2 Episode 2, starting at 1:06:16 as the camera drifts up the stairs. If the camera wasn't moving, it's harder to see.

It would be really great to have moving patterns as well, so making it as easy as possible to see.

One of the patterns I am really excited about is pixel aging. This will allow new OLED TVs to break-in before initial calibration. You want to run it for 100 hours. It may also help with image retention.

Ooh, if it does help with IR that will surely make a lot of people very happy!

[Stacey Spears]

Originally Posted by mrtickleuk

It's a lot more when rotating and zooming. I think it's because it forces the TV to use different pixels for the black/not-black transitions as the pattern moves. That is also combined with the TV's temporal dithering patterns.

There is a lot less flashing with LG's workaround/fix firmware than the original firmware, though.

Thanks, but to clarify, is this with a moving band as in that guy's video? It's the sweeping diagonal movement that really helps to show it. Much like many of the "watch scene X in episode Y" test show up the banding because the camera is moving across a statically-dark patch of scenery, and it's as that near-black scenery moves across the TV do you notice the banding.

I had not watched the video all the way through, I saw the static gray with streaks and stopped.

Thank you for the additional info on the flashing pattern too. It looks like there is a gray half and a black half for the background. Is one side better than the other?

[p5browne]

Do double layer BDs give you any more room ?
Reason I ask is, Windows 10 64 Bit, now requires these when one wants to burn the new updates. 32 Bit still single layer.

[Stacey Spears]

Originally Posted by p5browne

Do double layer BDs give you any more room ?
Reason I ask is, Windows 10 64 Bit, now requires these when one wants to burn the new updates. 32 Bit still single layer.

Yes, double the space. The current disc is a triple layer.

[mrtickleuk]

Originally Posted by Stacey Spears

I had not watched the video all the way through, I saw the static gray with streaks and stopped.

Aha. Are you persuaded, then?

Thank you for the additional info on the flashing pattern too. It looks like there is a gray half and a black half for the background. Is one side better than the other?

Yes, I would say that on my LG C8 with fw 5.10.40, the left side (with the grey background) shows the problem a bit worse than the right side which has a black background.

[mrtickleuk]

Originally Posted by mrtickleuk

Great thanks. I will try that at home over the weekend.

Unfortunately that dropbox link isn't working for me. Has it already expired?

[Stacey Spears]

Originally Posted by mrtickleuk

Unfortunately that dropbox link isn't working for me. Has it already expired?

No, this forum changes the link after a while for reasons I don't understand. It is why the menu mock-ups work for a little while and then stop. I will PM it to you. Seems your PM is disabled.

See if the spoiler tag works. You will have to copy and paste the link.

[Show spoiler]https://www.dropbox.com/s/iqbjh6c39ukmarj/2D_Quant_Cr_20_01_04.zip?dl=1

[iceman]

Originally Posted by Stacey Spears

No, this forum changes the link after a while for reasons I don't understand. It is why the menu mock-ups work for a little while and then stop. I will PM it to you. Seems your PM is disabled.

See if the spoiler tag works. You will have to copy and paste the link.

[Show spoiler]https://www.dropbox.com/s/iqbjh6c39ukmarj/2D_Quant_Cr_20_01_04.zip?dl=1

If you post dropbox links with "dl=0" it will work
("dl=1" when followed will be rewritten to a new temporary url that later on expires).

[mrtickleuk]

Originally Posted by Stacey Spears

No, this forum changes the link after a while for reasons I don't understand. It is why the menu mock-ups work for a little while and then stop. I will PM it to you. Seems your PM is disabled.

See if the spoiler tag works. You will have to copy and paste the link.

[Show spoiler]https://www.dropbox.com/s/iqbjh6c39ukmarj/2D_Quant_Cr_20_01_04.zip?dl=1

Thankyou! No idea why there was a PM problem as other people have sent me PMs before, and I only have 6 out of 100 in there . But I've got the link working now changing to dl=0, many thanks iceman, and I've saved it to my own dropbox for later

[Stacey Spears]

Originally Posted by iceman

If you post dropbox links with "dl=0" it will work
("dl=1" when followed will be rewritten to a new temporary url that later on expires).

Thank you for the info. I am not a fan of dl=0 as it takes you to dropbox vs. opening a save dialog right away.

[mrtickleuk]

I've tried the patterns. Very interesting, and I do see a little bit of the overshoot flashing actually!
Because it's so very dark, I'm struggling to make out what the colours should be. The Cr patterns seem more red at the bottom and more green at the top?
The Cb patterns - I can't tell

[Stacey Spears]

Originally Posted by mrtickleuk

I've tried the patterns. Very interesting, and I do see a little bit of the overshoot flashing actually!
Because it's so very dark, I'm struggling to make out what the colours should be. The Cr patterns seem more red at the bottom and more green at the top?
The Cb patterns - I can't tell

You need to really view it in a dark room.

Cr is a green to red transition. Cb is a yellow to blue transition. Its very low level, so it is why you need a very dark room. You posted about the original version of the pattern at AVForums. Did you actually take those photos? If so, how did you get them to show that way? Or rather, which version of the patterns did you use?

[Stacey Spears]

Originally Posted by mrtickleuk

Thankyou! No idea why there was a PM problem as other people have sent me PMs before, and I only have 6 out of 100 in there . But I've got the link working now changing to dl=0, many thanks iceman, and I've saved it to my own dropbox for later

I see the attached image, just tried again.

[mrtickleuk]

Originally Posted by Stacey Spears

You need to really view it in a dark room.

Yep, did, pitch black!

Cr is a green to red transition. Cb is a yellow to blue transition. Its very low level, so it is why you need a very dark room.

Thanks. It's just that it is so so dark, and the colours aren't so good on my C8 at those very low light levels, they were too dark for me to tell which colours they were.

EDIT: I think the problem is that it's only 505-519 on the vertical axis (but perhaps that is needed to see the effect)? This is almost the same colour.

It just looks like grey with no detectable chrominance.

On the HDR version of the red-green pattern, when it's at its brightest, I can JUST detect "that's a bit like red" and ditto for green.

I can see the colours are not grey, but I can definitely not tell which non-grey colours they are supposed to look like. There's just not enough saturation, or internal "bits" in the OLED T-con, or whatever. Even knowing it's supposed to be blue-to-yellow, I can't see either of those colours and couldn't tell which whether the top or the bottom should be blue.

I know this problem won't happen on the disc, but the first few seconds are taken up by the "HDR" piece of toast, then the next few seconds by a "this video does not support audio" which is massively bright. It's pretty much 10 seconds into each clip before I have the first chance to (try to) make out anything at all on the screen! And that's not taking into account time for my eyes to adjust after the searing-bright "toast" items have disappeared.
Encoding with very low bitrate silence, instead of no audio, would be a lot better

You posted about the original version of the pattern at AVForums. Did you actually take those photos? If so, how did you get them to show that way? Or rather, which version of the patterns did you use?

I don't think that's the correct link, which post # in the thread is it? On that page I only talk about how to display the TV's built in test patterns.
However, I think around 10th March 2019 I took pics just after I got the LG "fixed" firmware. If they are the ones I think you mean they were done in a dark room on my phone, but I do struggle taking visible pictures in the full dark on it.

Originally Posted by Stacey Spears

I see the attached image, just tried again.

Cool thanks. Found it. There's an option called "Receive Private Messages only from Contacts and Moderators" which I had ticked. Now it is un-ticked!

[Geoff D]

I concur with the Mr Man, bearing in mind that I'm watching on one of those filthy LCD thingies. I can see the gradations in the Cb HDR pattern but virtually no colour at all, whereas in the Cr I can also just detect that green to red transition. (This is watching in a darkened room with no other light source than the TV itself.)

[Stacey Spears]

Originally Posted by Geoff D

I concur with the Mr Man, bearing in mind that I'm watching on one of those filthy LCD thingies. I can see the gradations in the Cb HDR pattern but virtually no colour at all, whereas in the Cr I can also just detect that green to red transition. (This is watching in a darkened room with no other light source than the TV itself.)

When it comes to color. An LCD will produce better colors at high nit levels while OLED will produce better colors at low nit levels. A true RGB OLED will be superior in both, but that is not really an option.

[Stacey Spears]

Originally Posted by mrtickleuk

EDIT: I think the problem is that it's only 505-519 on the vertical axis (but perhaps that is needed to see the effect)? This is almost the same colour.

Yes, that is correct.

The vertical plane has Cb or Cr starting at 512 in the center, which would mean black and white. Then its +/- 7 moving up or down. On the Cb pattern, those are all 512 in Cr and visa versa.

In the horizontal plane it is Y ramping from 64 to some value (don't have it in front of me at the moment.) Only Y changes value as it ramps up and down. Cb and Cr stay the same.

[Geoff D]

Originally Posted by Stacey Spears

When it comes to color. An LCD will produce better colors at high nit levels while OLED will produce better colors at low nit levels. A true RGB OLED will be superior in both, but that is not really an option.

I know: https://forum.blu-ray.com/showthread...e#post15462303

I'm just saying that I'm seeing much the same thing on my janky old LCD as what mr tickle is on his OLED with those HDR Quant patterns, that one appears to have no colour at all while the other has a hint of it. Minus any flashing/'overshoot', natch.

[mrtickleuk]

Originally Posted by Stacey Spears

Yes, that is correct.

The vertical plane has Cb or Cr starting at 512 in the center, which would mean black and white. Then its +/- 7 moving up or down. On the Cb pattern, those are all 512 in Cr and visa versa.

In the horizontal plane it is Y ramping from 64 to some value (don't have it in front of me at the moment.) Only Y changes value as it ramps up and down. Cb and Cr stay the same.

Ok, many thanks. My head hurts when it comes to YCrCb, so thankyou for the numbers.
On the horizontal plane, it starts with 64 on the left and 90 on the right (so 77 would be the middle of those two).
the brightest point which I'll call the "top of the ramp", has 90 on the left and 116 on the right (so 103 would be the middle of those two).

I have found an online calculator to convert from YCrCb to RGB. If anyone knows of a better one, please shout! This one only works in ranges from 0-255, so I had to divide by 4 and then round some fractions, for the conversion to be possible.

It is here: http://www.picturetopeople.org/p2p/i...esultType=view

However, I'm confused with the results, I could have very easily made a mistake, so I will post here. If I am right, though, it could explain why I am not seeing any colours, because it's just a huge mass of greys! Also, it seems (going from the numbers coming back from the calculator) to be blue to green, and not blue-to-yellow.







EDIT: That should say TOP of ramp. Already edited this post too many times

I'm fully expecting it to be something around the conversion. But as I said, it's just so so dark and lacking in colour, you never know.

Finally I have to thank you again for this because only now do I finally understand what the Test Full Band "Quants3D" patterns are doing. They are static grids, with the "Y" component changing over time. Whereas the "Quants2D" patterns are just the static grids with a Y ramp from left-right and either Cr ramp or Cb ramp top to bottom, and I'm now inferring that the non-Cr or non-Cb component is a flat value (128?) for the whole grid in each case.