Sony introduces high-end ZD9 TV series with HDR

[Stacey Spears]

Originally Posted by Penton-Man

I’m a fan

I actually bought the Shogun Flame for the very reason you stated in that thread! I also like that they have put in support for various camera log modes to display HDR.

It is not perfect, but works pretty well. Hopefully they can improve it. Here is what it currently looks like with FW 7.11. What this graph is showing is that they are not tracking ST2084 as well as they could. They are coming in low. I measured 1300+ nits. I would say they are doing a better job than a lot of consumer HDR displays though.


I shared this file with others a few months ago. Not sure if everyone here has seen it. It is a prototype of a banding test pattern. This version is luminance only. You can put it on a USB drive and play it in most displays. The Samsung UHD BD player will also play it.

The image embedded here has been converted to 8-bit and will show banding in all boxes.

From left to right: 10-bit to 8-bit rounded, 10-bit to 8-bit dithered and 10-bit. It is a shallow gradient that rotates. So far I have only seen a few displays that appear to be true 10-bit end-to-end. These include the Sony 940C, Vizio R65 and a 2014 Panasonic (Don't recall the model number). All other consumer displays appear to be using dither.

For those that receive their Z9D first, you can try it yourself. I am also willing to share the uncompressed image sequence if anyone wants it.

The original version only had the 8-bit rounded and 10-bit. I added the dither when I saw the Samsung JS9500. The dither patch looks very much like the JS9500. The second version, which added dither, applied dither once and then rotated. The current version applies dither to each frame.

In future versions I will experiment with chroma, rotational speed and fading.

I will hopefully have a Z9D in a couple of weeks. I am more than happy to share my results, but understand I am not doing a "formal" review. I will simply provide raw data. If there are certain aspects people would like me to look at, feel free to ask.

[ray0414]

Originally Posted by Stacey Spears

Just to add to the image I had posted earlier. Here is a histrogram, wave form monitor and even the pixel values of one pixel on screen.


Here are the source frames as 16-bit DPX files.

The SDR have been mapped into HDR so they look like SDR when in HDR mode.

1. SDR 709
2. SDR 2020
3. HDR 500 nits 2020
4. HDR 1000 nits 2020
5. SDR 709 and HDR 1000 2020 in butterfly
6. The original R3D source frame. This way you can do your own SDR and HDR version if you don't trust 1-5. Log3G10 and REDWideGamutRGB were used as the starting point to create 1-5.

The above screen shots were taken using Scratch Play. http://www.assimilateinc.com/products/scratch-play/ Scratch Play is a great media viewer. It is a free version of their professional app. If you are unfamiliar with Scratch, you can read more about it.

Also planning to animate the still frame and encode as HEVC. There will be two versions. One that does

One of the HEVC clips is a starfield. I got the idea from the opening of Pacific Rim. Its been good clip to test FALD vs. Edge lit vs. OLED. There are four levels of stars. From very few stars to massive amounts of stars. On part 1, the LG OLED looks like a beautiful starfield while on the JS9500, you would not even know there was a starfield.

If others have ideas on what they would like to test the ZD9, or other, HDR displays, please describe it and I will try and create it.

Excellent stuff.

Can't wait to hear your impressions of the Z9D. Any plan to figure out a way to count the leds? Guesses are ranging from 800 up to 12,000 on the 65", but my personal guess is 2500 (35x70).

Have you seen the ks9800 yet? The word is that it is the only display that tone maps properly up to 4000 nits without clipping.

[Stacey Spears]

Originally Posted by Penton-Man

Liking the use of star fields to prove a point for a long time

Here is my attempt at a starfield. This was created in Adobe After Effects. Each starfield begins and ends with black. There are four levels of stars. The 4th has an absurd number and the encoder did not perform very well with it. I will try re-encoding later and if that does not clean it up, I will dump the 4th.

I consider this a prototype too. At the time I made this, Kris Deering had a little gathering to compare a Sony 4K (Full 4k, not UHD) and a JVC faux-UHD projector. I had noticed that the opening of Pacific Rim looked milky (Lacked contrast) on the Sony compared to the JVC. The bright scenes looked great on the Sony in the contrast ratio department.

The starfield turned out to be a good clip when looking at HDR displays. Really shows what an OLED can do. Because the stars do go to 940, I have set the HDR metadata to 10,000 nits.

In this graph, the smallest window we have is 1% (area)


We are in the process of creating a sub 1% area patch. The window will be just slightly larger than the opening of a Klein K80. I wish we could actually measure just a single pixel.

[Stacey Spears]

Originally Posted by ray0414

Have you seen the ks9800 yet? The word is that it is the only display that tone maps properly up to 4000 nits without clipping.

Not yet, but we will be measuring one for our paper. What does "tone maps properly" actually mean? Since there is no standard on how HDR displays should tone map it sounds like something a marketing department would say.

Originally Posted by ray0414

Any plan to figure out a way to count the leds?

I would like to find a way to quantify the quality of zone based systems. The Murideo generator has a thin moving white line. You can specify horizontal or vertical direction. We were watching it on one display. You could see zones turning on and off as it moved. It has given me some ideas, but you would mostly be forming subjective opinions from it. e.g. on this particular display my subjective opinion was that it was Crap!

[DanBa]

Originally Posted by ray0414

Have you seen the ks9800 yet? The word is that it is the only display that tone maps properly up to 4000 nits without clipping.

Is this tone mapping algorithm known hence validated by the mastering artists?

[DJR662]

I am still baffled by the fact this TV comes out so soon after the XD94, if I had prior knowledge of this I would have waited for the ZD9 to come out.

[ray0414]

Not sure if either of you (Stacy and DanaBa) have been following Vincent Teohs tests, but hes been doing cross comparing of DV and HDR10 and using his own methods to determine clipping on tvs.

In his original article, he was able to test DV vs HDR10 on an oled E6, and the oled is clipping details on both 1000&4000 nit mastered content.

http://www.hdtvtest.co.uk/news/dolby...1606214303.htm

As expected, Dolby Vision correctly maps to the display and all highlights and details are properly rendered, but show up "missing" when playing the HDR10 version. This includes an entire sun in the background, and the "bubble" that the characters are inside of in "Pan". He then cross compared to a KS9800 and found the samsung to be displaying all of the detail that the DV version was showing.

Vincents conclusions:

*Dolby Vision delivered a superior HDR image compared with HDR10 4K Blu-ray on the LG OLED55E6, displaying less highlight clipping and more accurate colours;

*The most effective method to recover the blown-out highlights in Ultra HD Blu-ray movies on the LG E6 was by lowering the [Contrast] on the source player; and

*A top-end 4K HDR LED LCD TV with high peak brightness and correct tone-mapping (for example the Samsung UE65KS9500) could present HDR10 UHD Blu-ray films in a manner that’s not inferior to Dolby Vision.


Vincent recently had his UK shootout, and the KS9800 was voted the best HDR tv, mostly in part because it was the only tv that didnt clip the content.

http://www.hdtvtest.co.uk/news/shoot...1608054331.htm (with detals of content used)

Samsung KS9800: properly tone mapping 1000 and 4000 nit mastered content.
Sony 940D: clipping all HDR10
Panasonic DX902: clipping 4000 nit mastered content
2016 oleds: clipping all HDR10


Itll be interesting to hear if Sony has corrected the tone mapping for 4000 nit content with the Z9D tv, otherwise all that extra brightness will be for nothing.


Edit: here's a picture of an lcd hdr tv next step to oled E6, the e6 is clipping the sun/details. Pretty sure lcd is a dx902. This is from vincents recent E6 review. He's been putting alot of tvs side by side and dojo g some great work. This scene is from kingsman.

[Penton-Man]

Originally Posted by Stacey Spears

...and REDWideGamutRGB were used as the starting point to create 1-5.

I missed that ^ when I skimmed your post yesterday. Sounds new (not to mention being a mouthful to say ). I assume you remain pleasantly rockin with a RED camera (which one now?) and all its latest accessories (color science).

[Penton-Man]

Originally Posted by Stacey Spears

I actually bought the Shogun Flame for the very reason you stated in that thread!....

We are but a small humble group in that thread.

[Penton-Man]

Originally Posted by Stacey Spears

Here is my attempt at a starfield.....

Nice . A piece of advice....if you ever decide to publish the star field part conjoined with the subjective impressions of that particular imagery by a group of observers in a standardized environment, best (and easiest) make sure that all your observers have at least 20/20 visual acuity…..and if less (20/20-2, 20/25, etc.), then have them a history of a recent perfectly normal visual field test (think a star field with a white background, rather than nighttime).

Otherwise, a knowledgeable editor could red flag the ‘ methods and materials’.

[DanBa]

Originally Posted by ray0414

Not sure if either of you (Stacy and DanaBa) have been following Vincent Teohs tests, but hes been doing cross comparing of DV and HDR10 and using his own methods to determine clipping on tvs.
...

I think every HDR TV comparison should include a reference mastering display.

In case of color volume mapping at the TV playback time, the TV should get mapping instructions (i.e. adaptation for consistent visual metadata) generated at the mastering time (i.e. validated by the filmmakers).

Without an adaptation for consistent visual protocol, a given TV manufacturer’s color volume mapping algorithm could not guess every filmmaker intent.

That is why SMPTE ST 2094 is being standardized as far as I understand.

HDR10 = ST 2084 + ST 2086
Dolby Vision "v2" = ST 2084 + ST 2086 + ST 2094-10




Which of the two following images do you prefer?



(What is HDR? - Technicolor HDR Series Workshop - YouTube)

In fact, you don’t have to choice, because these two images represent two different stories the filmmaker wants to tell you.

[Stacey Spears]

Originally Posted by Penton-Man

I missed that ^ when I skimmed your post yesterday. Sounds new (not to mention being a mouthful to say ). I assume you remain pleasantly rockin with a RED camera (which one now?) and all its latest accessories (color science).

It is brand new. They are calling it RWG for short. This thread on RU has more detail. Mike does nice job summarizing it.

I currently have a CF Weapon. I start shooting a doc on 12/5 with the Helium sensor. I have been waiting for my VV upgrade since NAB 2015. For optics, I am using the Otus lenses. The 85mm Otus makes a wonderful image for talking head interviews.

I have been working on image deconvolution to undo the blur caused by the OLPF in front of the sensor. Here is a sample of my decon.
1. Original image
2. Decon applied
3. A difference image so you can see the pixels touched

I would use the old Windows Image viewer. Open 1 and 2 into their own copy of image viewer. Go 1:1 and then alt+tab between them. Take a look at the signs on the trash can. Scratch Play also works and then you can just use the arrow keys to go between the two images.

If you follow the philosophy of the stills world with capture, creative and output sharpen, this would be capture sharpen. The motivation for decon came from Bart.

Also looking at anti-ringing image scaling. There is a paper on SSIM scaling along with several different scaling examples in this file. The source image, windmill.jpg, is extremely sharp and the standard SSIM is a bit too much. SsimAb looks subjectively nicer. The source image also used decon. I would like to find something similar in WA to test the scaling. I am focused on 2:1 in each direction so that I can go from 7680x4320 to 3840x2160. It is an easier ratio.

Take a look at the bilateral folder as well. I think it is a big improvement with chroma upsampling. There are three examples with the SMPTE bars. Catmull-Rom, Catmull-Rom with anti-ringing and bilateral. Then the red bar on yellow background is a big improvement.

The anti-ringing folder is the same anti-ringing used on the Catmrull-Rom chroma upsampling and the scaling.

All of the above samples, except deconvolution, were done using madVR. I started with the Raw Therapee deconvolution algorithm. I added an adaptive iteration feature and I changed it from working on L of Lab to Y of xyY. I believe decon should be done in linear light while traditional sharpening should be done in a perceptual space.

I did try decon in linear RGB, but it was much slower. I do want to revisit at some point. Since I am shooting on a CMOS sensor, I wonder if the amount of decon should be optimized for green differently than red and blue. There are a lot of cool stuff that can be done, sadly I don't have time to try everything.

I would certainly love feedback on decon, if anyone wants to take a look at the sample image. Same for the different scaling algorithms.

This post is already long enough.

[Stacey Spears]

Originally Posted by ray0414

Not sure if either of you (Stacy and DanaBa) have been following Vincent Teohs tests, but hes been doing cross comparing of DV and HDR10 and using his own methods to determine clipping on tvs.

Thank you for sharing! Do you happen to have the timestamp for that scene in Pan?

I think what was throwing me off was, "...the only display that tone maps properly" implies that there is a single proper way to tone map. e.g a published standard.

1. I need to purchase Pan. Someone else pointed out another thing to look at in Pan. I don't remember what it was at the moment.

2. Saying that it "tone maps properly" implies that there is a well defined spec on how to tone map, which there is not. This is one aspect of Dolby Vision that I really like.

3. I think it would have been better to say that it has superior tone mapping in that it does not clip while others do.

I was wondering why they had said the Samsung was the best HDR in their shootout. Are there other examples of this clipping besides Pan? The Amazon clip is a good one, but not something everyone has access to. I want to create a synthetic clip that showcases the same thing Pan does.

[ray0414]

Originally Posted by Stacey Spears

Thank you for sharing! Do you happen to have the timestamp for that scene in Pan?

I think what was throwing me off was, "...the only display that tone maps properly" implies that there is a single proper way to tone map. e.g a published standard.

1. I need to purchase Pan. Someone else pointed out another thing to look at in Pan. I don't remember what it es.

The bubble detail issue of Pan has a time stamp In the link covering the hdr shootout results (19:10 I think), as is a few other time stamps. The scene with the sun in Pan that is missing on the oleds is also time stamped in the 1st article I linked.


You're right, there is no standard to tone map. Vincent refers to samsungs tone mapping as "clever".

I had the js9500 and ks9800 side by side at my house, the difference was quite shocking in many areas: deepness of native blacks, color saturation, bloom control and off angle blooming too (though not perfect), also detail of highlights. Playing the same hdr demos side by side, the ks9800 was not also brighter with deeper blacks, but the detail was much more refined on the ks9800. For instance during the nasa hdr demo, there's an upclose shot of the sun with flames shooting out. On the js9500 the flames were more blown out but on the ks9800 you could see all the detail in the flames and control over the highlights with color having definition near the flames. This shows me improved tone mapping for hdr content over the js9500 being able to tender bright details and surrounding areas of highlights. I expect the sony Z9D to surpass the ks9800 in every way and be strong in the area and that the ks9800 is weak in.

[Penton-Man]

Originally Posted by ray0414

Vincent refers to samsungs tone mapping as "clever".

Li is a clever lady.

[ray0414]

Originally Posted by Penton-Man

Li is a clever lady.

And very pretty too! Wonder if she helped create the hdr tone mapping for samsung?

Penton have you seen the Z9D yet? Did you attend the video and audio center launch?

[Penton-Man]

Originally Posted by Stacey Spears

I currently have a CF Weapon. I start shooting a doc on 12/5 with the Helium sensor.

I see you’ve come a long way since the days of the ‘4K’ Red One (3.1 or 3.2K rez?). I hope you’ve accumulated some good memories with the journey.

Originally Posted by Stacey Spears

This post is already long enough.

Yes, I fear we’re drifting too much from the original topic of this thread, namely, discussing the Z9D. On that note, to tighten it up a bit, heads-up to Orange County folk....BB has ~ 50 of the 65 inchers in their Chino warehouse. Get ‘em while they’re a hot topic of discussion and share your impressions of the performance of the TV with others contemplating pulling the trigger.

[Penton-Man]

Originally Posted by ray0414

And very pretty too! Wonder if she helped create the hdr tone mapping for samsung?

Penton have you seen the Z9D yet? Did you attend the video and audio center launch?

1. she’s a tone mapping wizard….with patents pending to prove it.
2. Yes.
3. No.

[Stacey Spears]

Originally Posted by Penton-Man

Yes, I fear we’re drifting too much from the original topic of this thread, namely, discussing the Z9D.

As soon as I get one, I will try and post the same number of words on it to even the score.

Do we know if the performance between the 65 and 75 are similar?

Some questions for those that get them:
1. Is HDR disabled over HDMI be default? The Sony 940C and the Panasonic have done this. I am told for compatibility reasons.
2. Can you manually enable HDR like you can on the 940C?
3. Which apps are built-in that support HDR?
4. Is there a wall mount kit available?

[Penton-Man]

Originally Posted by Stacey Spears

As soon as I get one, I will try and post the same number of words on it to even the score.

Sounds excellent . Looking forward to it.

We will be even more intrigued if those words are exclusive (to Blu-ray.com).
Since thee earliest of times, people enjoy exclusive info.