Dolby Vision, including HDR10 conversion w/ DTM on Projectors? - Page 26 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 626Likes
Reply
 
Thread Tools
post #751 of 1392 Old 02-11-2020, 06:24 AM
Advanced Member
 
aerodynamics's Avatar
 
Join Date: Oct 2007
Posts: 613
Mentioned: 8 Post(s)
Tagged: 0 Thread(s)
Quoted: 311 Post(s)
Liked: 136
This might not be the best place to ask but thought I'd share anyway. I just bought a Chiq B5U UST projector. It's a newer, obscure Chinese model made by Changhong. I like the image and quality a lot. One surprising thing I noticed is when I connected my ATV 4k, it asked whether to switch to Dolby Vision mode. I selected it out of curiosity and nothing looked out whack. Then I noticed the available picture modes of the projector switched to "Dolby Vision Bright" and "Dolby Vision Dark" instead of the usual game/movie/PC, etc. When playing DV enabled content, the DV logo appears at the start and the picture does not look bad. Not great, but not bad, definitely not abnormal. I think the image I get from madVR tone mapping is much better. My question is, I was under the impression there are no projectors on the market that are capable of DV. However, the projector obviously recognizes it and the devs specifically wrote it into the menus. So what is happening here?
Archibald1 likes this.

Chiq B5U | 100” Snowhite UST Screen | Pioneer Elite LX-503 | 7.2.4 Surround | Sony Core Series LCR + Atmos | Cambridge Min 12 Side Surrounds | Monoprice Rear Atmos Surrounds | Bic V1020 x2
aerodynamics is offline  
Sponsored Links
Advertisement
 
post #752 of 1392 Old 02-11-2020, 06:34 AM
AVS Forum Special Member
 
thrillcat's Avatar
 
Join Date: Mar 2013
Location: Ames, Iowa
Posts: 1,607
Mentioned: 56 Post(s)
Tagged: 0 Thread(s)
Quoted: 695 Post(s)
Liked: 738
Quote:
Originally Posted by aerodynamics View Post
This might not be the best place to ask but thought I'd share anyway. I just bought a Chiq B5U UST projector. It's a newer, obscure Chinese model made by Changhong. I like the image and quality a lot. One surprising thing I noticed is when I connected my ATV 4k, it asked whether to switch to Dolby Vision mode. I selected it out of curiosity and nothing looked out whack. Then I noticed the available picture modes of the projector switched to "Dolby Vision Bright" and "Dolby Vision Dark" instead of the usual game/movie/PC, etc. When playing DV enabled content, the DV logo appears at the start and the picture does not look bad. Not great, but not bad, definitely not abnormal. I think the image I get from madVR tone mapping is much better. My question is, I was under the impression there are no projectors on the market that are capable of DV. However, the projector obviously recognizes it and the devs specifically wrote it into the menus. So what is happening here?


Likely not an authorized use. The fact that this off-brand Chinese projector is claiming Dolby Vision would likely be very interesting news for the folks at Dolby Labs.


Sent from my iPhone using Tapatalk

I love my iPhone, but it will never replace my turntable. I've never seen Titanic OR XXXXXXXXX. Alexa is a stripper name.

The Cinema 1858 Remodel Thread
thrillcat is offline  
post #753 of 1392 Old 02-11-2020, 03:22 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Here is the decoded Dolby VSB from the Sony A1 EDID file:

dovi_info
ver = 2
sup_2160p60hz = 1 // always assumed true
dm_version = 1
sup_yuv422_12bit = 1
sup_backlight_control = 0
sup_global_dimming = 0
backlt_min_luma = 3
tminPQ = 0
Interface = 0
tmaxPQ = 16 (0.500)
sup_10b_12b_444 = 0
Gx = 67 (0.263)
Gy = 176 (0.690)
Rx = 174 (0.682)
Bx = 38 (0.149)
Ry = 81 (0.318)
By = 15 (0.059)

Some notes:
My understanding is that 'ver 2' is limited to LLDV.
The encoded value for maxPQ has to be divided by 32, and the encoded values for RGB xy have to be divided by 255; though I'm not 100% certain on this.
The value for maxPQ of 0.500 is roughly equivalent to 100 nits.

I haven't mapped the RGB chomacity xy values into 2020 color space, but I'd assume this is the cause of the red tint issue people have mentioned.

byte 0 (0x45) in the data is version, dm_version, yuv422_12bit support and backlight control support
byte 1 (0x03) is global dimming support, min backlight luma, and min PQ
byte 2 (0x80) is interface and max PQ
byte 3 (0x86) is 10b 12b 444 support, and Gx
byte 4 (0x60) is Gy
byte 5 (0x76) is Rx and Bx
byte 6 (0x8f) is Ry and By
bmandra is offline  
Sponsored Links
Advertisement
 
post #754 of 1392 Old 02-11-2020, 04:04 PM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by bmandra View Post
Here is the decoded Dolby VSB from the Sony A1 EDID file:

I haven't mapped the RGB chomacity xy values into 2020 color space, but I'd assume this is the cause of the red tint issue people have mentioned.
Interesting. The chromaticity is very close to DCI-P3.

Code:
     A1    2020  709
Rx = 0.682 0.708 0.640
Ry = 0.318 0.292 0.330
Gx = 0.263 0.170 0.300
Gy = 0.690 0.797 0.060
Bx = 0.149 0.130 0.150
By = 0.059 0.046 0.060
I created the profiles based on these value, although haven't tested them thoroughly yet.

The A1 profile seems to remove the remaining red tint, at least when I tried out Mad Max Fury Road. Initial impression very positive! It removes the red tint and yet does not seem to desaturate the stronger colours (which V5 tends to do).
Attached Files
File Type: zip LLDV_A1.zip (38.9 KB, 205 views)
MOberhardt and atabea like this.

Last edited by Dominic Chan; 02-12-2020 at 06:37 AM.
Dominic Chan is offline  
post #755 of 1392 Old 02-11-2020, 05:52 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
Interesting. The chromaticity seems to be much close to Rec. 709 than to Rec. 2020.

Code:
     A1    2020  709
Rx = 0.682 0.708 0.640
Ry = 0.318 0.292 0.330
Gx = 0.263 0.170 0.300
Gy = 0.690 0.797 0.060
Bx = 0.149 0.130 0.150
By = 0.059 0.046 0.060
I created the profiles based on these value, although haven't tested them yet. Based on the chromaticity they don’t seem to tame the red as much.
It’s very possible my divisor is incorrect. I had to piece this together by combing through the public source code for the amlogic htmitx Linux driver. Hopefully this is enough of a start though that others can pick this up and help find anything I missed or got wrong. Having the ability for the community to edit this data is really the biggest open issue with this solution I think.
atabea and Archibald1 like this.
bmandra is offline  
post #756 of 1392 Old 02-11-2020, 06:46 PM
AVS Forum Special Member
 
atabea's Avatar
 
Join Date: Aug 2008
Posts: 1,791
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 873 Post(s)
Liked: 711
Quote:
Originally Posted by Dominic Chan View Post
Interesting. The chromaticity seems to be much close to Rec. 709 than to Rec. 2020.

Code:
     A1    2020  709
Rx = 0.682 0.708 0.640
Ry = 0.318 0.292 0.330
Gx = 0.263 0.170 0.300
Gy = 0.690 0.797 0.060
Bx = 0.149 0.130 0.150
By = 0.059 0.046 0.060
I created the profiles based on these value, although haven't tested them yet. Based on the chromaticity they don’t seem to tame the red as much.
EDIT: Take it back. The A1 profile seems to remove the remaining red tint, at least when I tried out Mad Max Fury Road. Initial impression very positive!

Hi Dominic, Am I correct in assuming that the LLDV-A1 profile you posted above is an update to the previous LLDV5 profile?

EDIT: I just noticed you edited your post. Thanks for sharing!!!
atabea is offline  
post #757 of 1392 Old 02-11-2020, 07:26 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
Interesting. The chromaticity seems to be much close to Rec. 709 than to Rec. 2020.

Code:
     A1    2020  709
Rx = 0.682 0.708 0.640
Ry = 0.318 0.292 0.330
Gx = 0.263 0.170 0.300
Gy = 0.690 0.797 0.060
Bx = 0.149 0.130 0.150
By = 0.059 0.046 0.060
I created the profiles based on these value, although haven't tested them yet. Based on the chromaticity they don’t seem to tame the red as much.
EDIT: Take it back. The A1 profile seems to remove the remaining red tint, at least when I tried out Mad Max Fury Road. Initial impression very positive! It removes the red tint and yet does not seem to desaturate the stronger colours (which V5 tends to do).
Just saw your edit, that’s great news! So now the question is should we modify the values in the EDID to match a 2020 calibration?
atabea likes this.
bmandra is offline  
post #758 of 1392 Old 02-11-2020, 08:19 PM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by bmandra View Post
Just saw your edit, that’s great news! So now the question is should we modify the values in the EDID to match a 2020 calibration?
No, you want the actual gamut of the display, which is way short of BT.2020.

BTW, I took a closer look at the A1 gamut (.), and it's very close to DCI-P3 (+), see the CIE diagram below. The JVCs have a very similar gamut, with the colour filter.



That's why the profile looks so good (meaning accurate) on my JVC!
Attached Thumbnails
Click image for larger version

Name:	A1_CIE.jpg
Views:	398
Size:	347.3 KB
ID:	2683894  
atabea and bmandra like this.

Last edited by Dominic Chan; 02-11-2020 at 08:37 PM.
Dominic Chan is offline  
post #759 of 1392 Old 02-11-2020, 08:30 PM
AVS Forum Special Member
 
atabea's Avatar
 
Join Date: Aug 2008
Posts: 1,791
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 873 Post(s)
Liked: 711
Thanks to Bmandra for providing the decoded Dolby VSB and thanks to Dominic for conducting the tests, creating the A1 profile and sharing with the community. You guys have advanced this little hack beyond expectations. Awesome work guys!!!.
atabea is offline  
post #760 of 1392 Old 02-11-2020, 08:41 PM
AVS Forum Special Member
 
MOberhardt's Avatar
 
Join Date: Jul 2006
Posts: 1,798
Mentioned: 44 Post(s)
Tagged: 0 Thread(s)
Quoted: 1310 Post(s)
Liked: 938
Can someone explain what these LLDV profiles .prof files are that Dominic has posted? Are they something you load into the HDFury?
MOberhardt is offline  
post #761 of 1392 Old 02-11-2020, 09:02 PM
AVS Forum Special Member
 
atabea's Avatar
 
Join Date: Aug 2008
Posts: 1,791
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 873 Post(s)
Liked: 711
Quote:
Originally Posted by MOberhardt View Post
Can someone explain what these LLDV profiles .prof files are that Dominic has posted? Are they something you load into the HDFury?
Dominic'a custom color profiles as well as his custom gamma curves are created specifically for, and can only be uploaded to, JVC projectors. None of these files can or should be uploaded to the HDfury devices.
MOberhardt likes this.
atabea is offline  
post #762 of 1392 Old 02-11-2020, 10:29 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Here is a quick and dirty Google Sheet to calculate new values for the DV string based upon calibration data: https://docs.google.com/spreadsheets...it?usp=sharing

I gave it public read access, but anyone should be able to download a copy.

In my case, the calibration numbers for my PJ are:
Gx - 0.3077
Gy - 0.6533
Rx - 0.6610
Ry - 0.3381
Bx - 0.1552
By - 0.0604

So my new DV string would be: "eb:01:46:d0:00:45:03:80:9c:4e:50:b7"

Will test this tomorrow.
vollans likes this.
bmandra is offline  
post #763 of 1392 Old 02-12-2020, 07:01 AM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
So I think the divisor I was using for tMaxPQ was incorrect. Took another look at it this morning and I think a tMaxPQ value of 16 is closer to 0.727 (or 1000 nits). This means a sensible value to use in the DV string for our PJs is either 58 (roughly 100 nits) or 60 (about 200 nits).



Default DV string at 100nits: "eb:01:46:d0:00:45:03:58:86:60:76:8f"

Default DV string at 200nits: "eb:01:46:d0:00:45:03:60:86:60:76:8f"



If anyone tries these out please let us know if there is any improvement.
bmandra is offline  
post #764 of 1392 Old 02-12-2020, 07:07 AM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by bmandra View Post
So I think the divisor I was using for tMaxPQ was incorrect. Took another look at it this morning and I think a tMaxPQ value of 16 is closer to 0.727 (or 1000 nits). This means a sensible value to use in the DV string for our PJs is either 58 (roughly 100 nits) or 60 (about 200 nits).



Default DV string at 100nits: "eb:01:46:d0:00:45:03:58:86:60:76:8f"

Default DV string at 200nits: "eb:01:46:d0:00:45:03:60:86:60:76:8f"



If anyone tries these out please let us know if there is any improvement.
For projectors people usually apply a multiplier to the PQ curve (typically 4 to 5). That's why the Panasonic UB820 "Projector / Low Luminance Display" is 500 nits, even though most projector produce about 100-150 nits peak white.
Dominic Chan is offline  
post #765 of 1392 Old 02-12-2020, 07:17 AM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
For projectors people usually apply a multiplier to the PQ curve (typically 4 to 5). That's why the Panasonic UB820 "Projector / Low Luminance Display" is 500 nits, even though most projector produce about 100-150 nits peak white.
Default DV string at 500nits: "eb:01:46:d0:00:45:03:70:86:60:76:8f"
bmandra is offline  
post #766 of 1392 Old 02-12-2020, 10:42 AM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by bmandra View Post
Default DV string at 500nits: "eb:01:46:d0:00:45:03:70:86:60:76:8f"
tMaxPQ doesn't seem to do anything on the ATV4K. I can see an obvious difference between HDR10 and DV, but not between the two DV (Default/80 vs 500/70).

For the EDID I used a HEX editor which did not update the checksum, so the Linker gave an error message. However, the EDID still loaded.
Attached Thumbnails
Click image for larger version

Name:	DV500_0004.jpg
Views:	48
Size:	858.9 KB
ID:	2684012   Click image for larger version

Name:	DV_0003.jpg
Views:	48
Size:	860.0 KB
ID:	2684014   Click image for larger version

Name:	HDR10_0001.jpg
Views:	42
Size:	832.1 KB
ID:	2684016  
Dominic Chan is offline  
post #767 of 1392 Old 02-12-2020, 10:54 AM
AVS Forum Special Member
 
markswift2003's Avatar
 
Join Date: Nov 2006
Location: UK, Europe.
Posts: 1,439
Mentioned: 40 Post(s)
Tagged: 0 Thread(s)
Quoted: 980 Post(s)
Liked: 686
Quote:
Originally Posted by bmandra View Post
Here is the decoded Dolby VSB from the Sony A1 EDID file:

dovi_info
ver = 2
sup_2160p60hz = 1 // always assumed true
dm_version = 1
sup_yuv422_12bit = 1
sup_backlight_control = 0
sup_global_dimming = 0
backlt_min_luma = 3
tminPQ = 0
Interface = 0
tmaxPQ = 16 (0.500)
sup_10b_12b_444 = 0
Gx = 67 (0.263)
Gy = 176 (0.690)
Rx = 174 (0.682)
Bx = 38 (0.149)
Ry = 81 (0.318)
By = 15 (0.059)

Some notes:
My understanding is that 'ver 2' is limited to LLDV.
The encoded value for maxPQ has to be divided by 32, and the encoded values for RGB xy have to be divided by 255; though I'm not 100% certain on this.
The value for maxPQ of 0.500 is roughly equivalent to 100 nits.

I haven't mapped the RGB chomacity xy values into 2020 color space, but I'd assume this is the cause of the red tint issue people have mentioned.

byte 0 (0x45) in the data is version, dm_version, yuv422_12bit support and backlight control support
byte 1 (0x03) is global dimming support, min backlight luma, and min PQ
byte 2 (0x80) is interface and max PQ
byte 3 (0x86) is 10b 12b 444 support, and Gx
byte 4 (0x60) is Gy
byte 5 (0x76) is Rx and Bx
byte 6 (0x8f) is Ry and By
Very nice indeed! Shame there's not a thumbs up emoji here.
Archibald1 likes this.

Zidoo Z9S | Nvidia Shield Pro | HDFury Vertex² | 32TB mirrored storage
Denon AVC-X8500H | JVC DLA X7000 | Screen Research 9ft multi aspect ClearPix 2 ISF screen
3 x PMC IB2S fronts | 4 x PMC Wafer 2 rears | 4 x PMC Wafer 1 ceiling | 2 x M&K SS550 THX height | Rel Stentor III
A pair of slippers, a comfy sofa and a glass of 16 year old single malt
markswift2003 is offline  
post #768 of 1392 Old 02-12-2020, 11:01 AM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
tMaxPQ doesn't seem to do anything on the ATV4K. I can see an obvious difference between HDR10 and DV, but not between the two DV (Default/80 vs 500/70).

For the EDID I used a HEX editor which did not update the checksum, so the Linker gave an error message. However, the EDID still loaded.
I'm using the Deltacast E-EDID editor to make changes here, but haven't had a chance to test any of these yet on my Diva. With the 2 DV pics I do see a difference in the blue, but hard to tell if thats just differences in the image itself.

I just got an Integral 2 today and noticed that with the latest firmware it's using a DV version 1 string that shows the decoded values, but the string isn't directly editable. I updated the code I'm using to decode these blocks to support v1 and get the same results as the Integral gui. Whats interesting is that with version 1 the block defines a min max luminance vs min max PQ. I have a feeling that my NITS<->PQ calculation is off...

Also, I did find a small error in the colorimetry conversion for the version 2 block. Here are the corrected values from the default A1 block:

_Gx = 0.262
_Gy = 0.688
_Rx = 0.680
_Bx = 0.148
_Ry = 0.316
_By = 0.059

Last edited by bmandra; 02-12-2020 at 11:09 AM.
bmandra is offline  
post #769 of 1392 Old 02-12-2020, 11:09 AM
AVS Forum Special Member
 
Join Date: Dec 2015
Posts: 1,156
Mentioned: 11 Post(s)
Tagged: 0 Thread(s)
Quoted: 783 Post(s)
Liked: 404
ok this was fun when it began a month back now you guys are way over my head with this stuff.. what are these numbers for what are you guys doing.. sounds fun but way beyond me
Superman2 and Archibald1 like this.
jorgebetancourt is offline  
post #770 of 1392 Old 02-12-2020, 11:25 AM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by bmandra View Post
With the 2 DV pics I do see a difference in the blue, but hard to tell if thats just differences in the image itself.
You’re right. I was only looking at white clipping and didn’t notice the difference in blue. Actually even for white there’s a small difference in luminance. I will compare them more closely.
Dominic Chan is offline  
post #771 of 1392 Old 02-12-2020, 11:29 AM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by jorgebetancourt View Post
ok this was fun when it began a month back now you guys are way over my head with this stuff.. what are these numbers for what are you guys doing.. sounds fun but way beyond me
The basic problem with this hack was always that it simply used the EDID values from a Sony A1 OLED to trick sources into supporting player led Dolby Vision, but nobody had any details on what exactly the data in the Dolby EDID block did. Now we have the ability to decode the raw block that the HD Fury device inserts into the EDID sent to the source and hopefully can modify it in ways that will improve the overall result.

There are essentially 2 things worth modifying:

1. The colorimetry values which represent the coordinates for peak red, green, and blue within the color space. If your display is calibrated and you know or can measure these values, then the DV block in the EDID should be able to be updated to ensure accurate color representation (ie. no red tint). Alternatively, if we know the values the source is using for peak red, green, blue then a custom curve can be made for the projector/display to compensate for any color shift which Dominic has already done.

2. The max luminance of the projector or the display. The Sony A1 TV is a 1000nit display, so when the source sees its value in the EDID block it tone maps to that peak value. With projectors our peak nits is much lower, and the hope is that by putting a more sensible value in the DV EDID block the source will tone map more appropriately.
bmandra is offline  
post #772 of 1392 Old 02-12-2020, 11:52 AM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
You’re right. I was only looking at white clipping and didn’t notice the difference in blue. Actually even for white there’s a small difference in luminance. I will compare them more closely.
Dominic,
Here's 2 other strings you could try. Both use the same colorimetry as the original A1 string and should be set for 350nits.

DV version 1: "ee:01:46:d0:00:26:0a:09:00:ae:51:43:b0:26:0f"
DV version 2: "eb:01:46:d0:00:45:03:28:86:60:76:8ff"


Edited to fix error in v2 block

Last edited by bmandra; 02-12-2020 at 12:23 PM.
bmandra is offline  
post #773 of 1392 Old 02-12-2020, 11:53 AM
AVS Forum Special Member
 
Join Date: Dec 2015
Posts: 1,156
Mentioned: 11 Post(s)
Tagged: 0 Thread(s)
Quoted: 783 Post(s)
Liked: 404
Quote:
Originally Posted by bmandra View Post
The basic problem with this hack was always that it simply used the EDID values from a Sony A1 OLED to trick sources into supporting player led Dolby Vision, but nobody had any details on what exactly the data in the Dolby EDID block did. Now we have the ability to decode the raw block that the HD Fury device inserts into the EDID sent to the source and hopefully can modify it in ways that will improve the overall result.



There are essentially 2 things worth modifying:



1. The colorimetry values which represent the coordinates for peak red, green, and blue within the color space. If your display is calibrated and you know or can measure these values, then the DV block in the EDID should be able to be updated to ensure accurate color representation (ie. no red tint). Alternatively, if we know the values the source is using for peak red, green, blue then a custom curve can be made for the projector/display to compensate for any color shift which Dominic has already done.



2. The max luminance of the projector or the display. The Sony A1 TV is a 1000nit display, so when the source sees its value in the EDID block it tone maps to that peak value. With projectors our peak nits is much lower, and the hope is that by putting a more sensible value in the DV EDID block the source will tone map more appropriately.
You guys are freaking animals keep it going.. So I'll wait on the sidelines till u guys do your thing then I'll test it. I'm super happy with the first gen DV version.. next up DV 1.1 update

Sent from my SM-G930T using Tapatalk
BobDyl, atabea, Archibald1 and 1 others like this.
jorgebetancourt is offline  
post #774 of 1392 Old 02-12-2020, 01:48 PM
AVS Forum Special Member
 
Join Date: Dec 2014
Location: Mississauga, ON, Canada
Posts: 8,112
Mentioned: 172 Post(s)
Tagged: 0 Thread(s)
Quoted: 5827 Post(s)
Liked: 2340
Quote:
Originally Posted by Dominic Chan View Post
tMaxPQ doesn't seem to do anything on the ATV4K. I can see an obvious difference between HDR10 and DV, but not between the two DV (Default/80 vs 500/70).

For the EDID I used a HEX editor which did not update the checksum, so the Linker gave an error message. However, the EDID still loaded.
The 200 nits EDID seems to give comparable brightness as HDR10. 500 nits gives better highlight details at the expense of overall brightness.
Attached Thumbnails
Click image for larger version

Name:	LLDV500_0015.jpg
Views:	71
Size:	711.2 KB
ID:	2684096   Click image for larger version

Name:	LLDV200_0016.jpg
Views:	72
Size:	766.8 KB
ID:	2684098   Click image for larger version

Name:	HDR10_0014.jpg
Views:	71
Size:	762.7 KB
ID:	2684100   Click image for larger version

Name:	DSC_0023-2.jpg
Views:	292
Size:	590.5 KB
ID:	2685446  

Last edited by Dominic Chan; 02-15-2020 at 05:45 PM.
Dominic Chan is offline  
post #775 of 1392 Old 02-12-2020, 01:49 PM
AVS Forum Special Member
 
atabea's Avatar
 
Join Date: Aug 2008
Posts: 1,791
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 873 Post(s)
Liked: 711
@bmandra , @Dominic Chan , @claw , @zombie10k and @Mark Swift will have this all figured out in short order
atabea is offline  
post #776 of 1392 Old 02-12-2020, 02:39 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by Dominic Chan View Post
The 200 nits EDID seems to give comparable brightness as HDR10. 500 nits gives better highlight details at the expense of overall brightness.
OK, so we're definitely moving in the right direction here.

Here is how to change the max luminance value using Excel:

DV v2 block: =DEC2HEX(BITLSHIFT(_VALUE_, 3))
DV v1 block: =DEC2HEX(BITLSHIFT(_VALUE_, 1))

_VALUE_ is 16 decimal for 1000nits (from the original A1 string). I think my original calculations for 200 and 500 were incorrect though, and believe that a value of 5 decimal is equivalent to 350nits though I can't be 100% sure. The v1 and v2 blocks have slightly different definitions, so it's not clear if internally they are treated the same by the source. The v1 block defines tmaxLUM while the v2 block defines tmaxPQ.

So 1000nits is 80 hex for DV v2 and 20 hex for v1. Similarly, 350nits is 28 hex in v2 and 0a hex in v1. As always I could be off slightly on some of the numbers.

Here is a simplified description of the bytes in the v1 block:
data[0] = 0xEE; // length
data[1] = 0x01; // always 1
data[2] = 0x46; // IEEE OUI
data[3] = 0xD0; // IEEE OUI
data[4] = 0x00; // IEEE OUI
data[5] = 0x26; // dm_ver, yuv 422 12bit support, 2160p60hz support
data[6] = 0x0a; // tmaxLUM, global dimming
data[7] = 0x09; // tminLUM, colorimetry
data[8] = 0x00; // Reserved always 0
data[9] = 0xa9; // Rx
data[10] = 0x57; // Ry
data[11] = 0x4f; // Gx
data[12] = 0xa7; // Gy
data[13] = 0x28; // Bx
data[14] = 0x0f; // By

And the same for a v2 block:
data[0] = 0xEB; // length
data[1] = 0x01; // always 1
data[2] = 0x46; // IEEE OUI
data[3] = 0xD0; // IEEE OUI
data[4] = 0x00; // IEEE OUI
data[5] = 0x45; // dm_ver, yuv422 12bit, backlight control
data[6] = 0x03; // tminPQ, backlight min luma, global dimming
data[7] = 0x28; // tmaxPQ, interface
data[8] = 0x86; // Gx
data[9] = 0x60; // Gy
data[10] = 0x76; // Rx, Bx
data[11] = 0x8F; // Ry, By
atabea and jorgebetancourt like this.
bmandra is offline  
post #777 of 1392 Old 02-12-2020, 05:17 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Tested a bunch of strings with different RGB values to match my calibration and AppleTV didn’t seem to like them. At least we know what the values are and can create an appropriate profile. Changing maxPQ does seem to have some impact but not a huge difference, need to play with that a bit more.
bmandra is offline  
post #778 of 1392 Old 02-12-2020, 08:27 PM
Newbie
 
Join Date: Jan 2020
Posts: 12
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 10 Post(s)
Liked: 1
Quote:
Originally Posted by bmandra View Post
Tested a bunch of strings with different RGB values to match my calibration and AppleTV didn’t seem to like them. At least we know what the values are and can create an appropriate profile. Changing maxPQ does seem to have some impact but not a huge difference, need to play with that a bit more.
I tried a custom string calculated from your Google Docs w/ RGB values for my projector and it worked. However, it would not work at 4k, but did work when I chose 1080 Dolby Vision in the AppleTV 4k... I am trying this with a new Vertex 2 I got today. I had previously been using an Integral 2 with the Sony A1 string just fine at 4k. The default Sony string didn't work for me on the Vertex 2 at 4k however, so I am not sure what's going on... Will try throwing the Integral at the end of the HDMI cable at my projector to rule out a weak signal or something...

Custom string I tried:

eb:01:46:d0:00:45:03:80:8C:5E:67:A2

tMaxPQ 0.727 80
Gx 0.2752 8C
Gy 0.6867 5E
Rx 0.6745 67
Ry 0.3255 A2
Bx 0.1512 7
By 0.0687 10

Also tried a different max PQ (100 nits?) one and I see the same thing. 1080p works but not 4k:

eb:01:46:d0:00:45:03:58:8C:5E:67:A2
ordivergens is offline  
post #779 of 1392 Old 02-12-2020, 10:07 PM
Member
 
Join Date: Jan 2020
Posts: 47
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 39 Post(s)
Liked: 32
Quote:
Originally Posted by ordivergens View Post
I tried a custom string calculated from your Google Docs w/ RGB values for my projector and it worked. However, it would not work at 4k, but did work when I chose 1080 Dolby Vision in the AppleTV 4k... I am trying this with a new Vertex 2 I got today. I had previously been using an Integral 2 with the Sony A1 string just fine at 4k. The default Sony string didn't work for me on the Vertex 2 at 4k however, so I am not sure what's going on... Will try throwing the Integral at the end of the HDMI cable at my projector to rule out a weak signal or something...

Custom string I tried:

eb:01:46:d0:00:45:03:80:8C:5E:67:A2

tMaxPQ 0.727 80
Gx 0.2752 8C
Gy 0.6867 5E
Rx 0.6745 67
Ry 0.3255 A2
Bx 0.1512 7
By 0.0687 10

Also tried a different max PQ (100 nits?) one and I see the same thing. 1080p works but not 4k:

eb:01:46:d0:00:45:03:58:8C:5E:67:A2
I think it has to do with automix having to remove bits from the original EDID to make the dv block fit. Have to test more to be sure though.
bmandra is offline  
post #780 of 1392 Old 02-12-2020, 11:39 PM
AVS Forum Special Member
 
claw's Avatar
 
Join Date: Jun 2001
Location: L'Etoile du Nord
Posts: 3,452
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 2736 Post(s)
Liked: 2580
Quote:
Originally Posted by Dominic Chan View Post
No, you want the actual gamut of the display, which is way short of BT.2020.

BTW, I took a closer look at the A1 gamut (.), and it's very close to DCI-P3 (+), see the CIE diagram below. The JVCs have a very similar gamut, with the colour filter.

That's why the profile looks so good (meaning accurate) on my JVC!
I loaded your two new profiles. I am not sure the no-filter version is disengaging the filter. But no matter as the filter version works really well.

I played a couple of ATV4K DV titles I am very familiar with; The Morning Show and Dickenson. I didn't realize how much red was still present until I switched to your new profile. Skin tones were especially good (Hailee Steinfeld, Jennifer Aniston, Reese Witherspoon, and Gugu Mbatha-Raw).

I then played the HDR10 version of Baby Driver using Amazon Prime App on the ATV4K. Again, very good result.

I used the default A1 EDID with a 1000 nit Arve curve. I don't see the need to try to lower the target nit value; it would just mean that we might need new custom curves to match. As an example, the 500 and 350 nit settings in the UB9000 had matching 500 and 350 nit auto-tone mapping profiles/curves in the new JVC projectors. Perhaps those with large screens and are nit starved might benefit from lowering the nit target but I have plenty of brightness.

Now some who have been praising the LLDV workaround might be disappointed that your new profile might appear to reduce some color saturation at first look compared to when using the JVC BT2020 color profile. But in reality the result is much more natural looking.

I think that your idea to create a JVC color profile based on the Sony A1 values rather than to try to change the A1 DV string values was a bit of genius. Unfortunately those with non-JVC displays might not be able to replicate this result so easily since they might not have the tools or know-how to create a custom color profile. So they will have to hope that experimenting with the DV string values can produce similar results.

Those with JVC projectors, I do recommend that you load Dominic's new A1 color profile (filter version) and try it combined with the A1 LLDV HDfury custom EDID in the same manner that we have done up to now. Without trying to adjust the DV string values. I can almost guarantee you will be pleased with the result.
atabea, mavang, Nalleh and 3 others like this.

CJ
JVC RS500 | LG B7A OLED | Denon X6400H/X4200W | Panasonic UB820 | Two Oppo 203 | Samsung K8500 | Apple TV 4K | HDfury Diva/Vertex/Linker/Integral | MadVR with EVGA RTX-2080 Ti

Last edited by claw; 02-13-2020 at 12:39 AM.
claw is offline  
Sponsored Links
Advertisement
 
Reply High Dynamic Range (HDR) & Wide Color Gamut (WCG)

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off