AVS Forum banner

HDR10 test patterns set

338K views 445 replies 82 participants last post by  Corplex 
#1 · (Edited)
HDR-10 calibration and test patterns set


Clip art Font Graphics Symbol Logo
Mehanik HDR10 calibration and test patterns set


HDR-10 calibration and test patterns set is designed and created by group of video enthusiasts. Main author of test patterns: mehanik99. Calibration patterns by me.
It's free and provided as is.

Feedback is welcome!

Donwload links:










List of HDR10 video test patterns

00. HDR-10 Calibration Patterns (CalMAN, ChromaPure and HCFR).

01. Black level setup

02. White&Color clipping
2.1. 10% sized windows
2.2. Extra metadata combinations (MaxCLL/FALL/MDL 0/1000/4000/10000)
03. Grayscale
3.1. Grayscale ramps
3.2. Grayscale steps
3.3. 10bit tests
04. Color reproduction
4.1. Color Ramps
4.2. Color Ramps Saturation
4.3. Linearization
4.4. Steps
4.5. Single colors
05. Chroma sharpness
5.1. Chroma sharpness&upsampling tests
5.2. SDR version
06. Aspect Ratio&Sharpness
6.1. Gray field filling
6.2. Pixel field
6.3. Gray field + pixel stripes
07. Frame Rates (including 23.976p) and judder

08. Audio Delay (central channel + 5.1) & Bitstream
8.1. SDR version
09. Test footage: sRGB (SDR) to HDR converted photos

10. ABL&Tonemaping

--


Additional test patterns:
- 4:4:4 [4K HDR10] chroma test for HTPC

- 10bit 1024 grayscale calibration steps for HDR10


--
Vladimir
 
See less See more
1
#2 ·
Uploaded patterns set to Google Drive and updated download links in first post:
https://drive.google.com/drive/folders/1m4IBq0euAxamL9ePgfdFuf8_5nLcRwHA

It's basically same files as on Yandex.Disk but Google Drive may be more convenient in some cases. Be aware that pressing "download all" button takes some time and google splits files into several different zip archives with 2Gb max size (currently - 5). After downloading just unzip all (currently 5) archives into same folder.
 
#4 ·
Any suggestions, advices, ideas, complaints?
 
#5 · (Edited)
As someone who is still learning calibration, I appreciate the manual that shows what to look for with some of the patterns as well as "bad" examples. I have a collection of disks (AVS, Ted, Ryan) that have dozens of various patterns and I often don't know what I am supposed to be looking for.

Some guidance for the test footage would also be helpful. I often don't find the test footage that helpful because I don't think I can spot a small error in the red color in a bowl of cherries. ; )

If there were a few examples with guidance that would really help. For example, have an example image for near black with instructions such as "look for the detail on the black button on the jacket" etc. Otherwise, I find myself just flipping through the patterns without much of a clue. ; )

I have seen the website that goes through a few Casino Royale scenes, but I've never seen a more comprehensive post-calibration, color reproduction workflow that would take someone through the 10-15 or so patterns / test footage they should check with guidance around what to look for (both good and bad). It would be great to eventually get to this point.

Thanks much for all the effort. I hope this helps.
 
#7 ·
Makes perfect sense to me and I'd love such a workflow with real content myself!

It's easier to add a description for synthetic tests as they are designed to show specific aspects of the image in the first place. But with real content it's hard to predict what could be broken. I think that obvious rule of thumb is: image shouldn't contain visible artefacts/image flaws: banding, clipping, posterization, etc.
Current version of test footage is based on public photos (in sRGB color space) which were converted to HDR10 format. sRGB originals are in IMG directorys so you can compare two version with each other. Don't expect a perfect match as HDR10 conversion has shifted both luminance and colors and most certainly TV settings/calibration are different but both images at least should look somehow alike.

Regarding the 10 bit test pattern: I would exchange the "good" example in the manual, because in fact it shows a bad one and is misleading. MadVR shows a fine smooth ramp with this test pattern and 8 Bit dithering. There is no need to show an "almost correct" one. :)
I've already though that this screenshot is not very illustrative. I think that some overexposure could help and make it more clear of what is expected to be seen. Also will add a note about 8bit+dithering.
 
#16 ·
#17 ·
Updated HDR10 test footage!
Direct link
(or can be found in 09. Test Footage directory)
Thankyou so much for this!

I do have a question actually. If I wanted to measure colour accuracy in HDR mode I obviously can't use fully saturated colours (100% saturation) since none of us have a panel which can do full BT2020. Also, there's no point trying to measure colours at 100% luminance either because of tone-mapping.

Therefore, with Calman the recommendation is to measure 50% saturation, 50% luminance colours. The question is which (if any) of your colour sets are 50% saturation, 50% luminance? in your section you have "04. Colors\05. Single colors"

but (a) I do not know what percentage luminance each of them are. It's surprisingly difficult to find a PQ lookup table :(
 
#20 ·
Thanks so much for this!

I will be using these with HCFR to calibrate my Panasonic OLED (EZ950) with HDR10. Would anyone be able to confirm if all I need are the MaxCLL 1000nit greyscale patterns? I'm not really sure what I am meant to use regarding nits, but I doubt my Panasonic would reach 4000 let alone 10K.

Thanks in advance!
 
#21 ·
Depending on MaxCLL/MaxFall meta TV could use different tone-mapping algorithms. Check out EZ950 review from hdtvtest, it contains PQ EOTF tracking graphs for different meta. My personal understanding is that as a base calibration starting point it is better to stick with HDR10 meta MaxCLL/MaxFall = 1000/400 as it's more often used in real content.

For CMS/color calibration you'll also need at least color saturation steps and color checker patterns to control your calibration. And, luckily, encoding of our first beta version of HDR10 calibration patterns for HCFR is almost done! I'll post a link in a few hours. For starters it will include: grayscale (10%, 5%, LG OLED 2016 and 2017 steps), near black&white, color primaries (Rec2020 and P3 in Rec2020), saturation 25% steps (Rec2020 and P3 in Rec2020), color checker 'GCD Classic'.
 
#22 · (Edited)
Exited to announce our first release of
HDR10 Calibration Patterns

Donwload link: HDR10 calibration patterns (check first post for latest version)

Current version includes patterns for HCFR 3.5.1.4 HDR-10 calibration workflow:
- grayscale (5%, 10%, LG OLED 2016 and 2017 steps)
- primaries and secondaries (Rec2020, P3 in Rec2020, Rec709 in Rec2020)
- near black (0-20%, 2% step)
- near white (96-100%, 2% step)
- color saturation (25% steps, Rec2020, P3 in Rec2020)
- color checker: GCD Classic

Patterns characteristics:
- MaxCLL: 1000 cd/m2
- MaxFALL: 400 cd/m2
- pattern size: 10%.
- duration: 2min (text label stays on screen for 10 seconds)

For combined version of patterns following timing is used: 5 seconds pattern with text label, 6 seconds pattern w/o text label, 4 seconds black screen cooldown.
Feel free to propose your variant of timing - it's quite easy to create custom versions (actually I'm thinking of creating separate directory for several combined versions with different timings).

All patterns are generated fully automatically using configuration text files with RGB codes or xyY coordinates. So it's easy to expand patterns set and create any custom workflows.
Comments and ideas are welcome!

 
#23 · (Edited)
Thanks so much! This is very much appreciated! Is it ok to use these patterns with Calman?

What I really need for measuring with Calman is BT.2020 50% saturation, 50% stimulus (about 94 nits) patterns, for RGBCMY.

In the directory "04. Saturation\01. Rec.2020 25% steps", I understand that the files with 50% in the names (eg "19-Rec2020-Yellow-100%.mp4") are 50% saturation, good. But what percentage is the stimulus of these files, please? If they were 50% stimulus, that is perfect. If not, would it be possible to have a set which are 50% stimulus please? (and white? :D)
 
#24 ·
It's 50% Amplitude (94.3 nits white).
E.g. for Rec2020 50% Red xyY target is 0.511320 0.310414 24.309546 (RGB 8 bit PC = 115 82 82)

Can you please post some Calman xyY targets and RGB codes (8 bit PC range)?
Calman version I do have (2016) uses significantly different xyY target for BT.2020 Red 50% Sat 50% Lum = 0,5082 0,3107 40,8259 (RGB 8 bit PC = 128 93 93)
 
#34 ·
@mrtickleuk
My earlier comment was regarding HCFR patterns and what I basically meant - you can't use them with Calman.

Archive I've posted later (-> Calman CMS ) is for Calman specifically and targets are exactly the same as you've posted. Also there is a text label below pattern (disappears after 10 seconds) and you can see if RGB triplet matches Calman expectation, see screenshoot in attach.
 

Attachments

#35 ·
I see. Those were HCFR numbers, and you wanted to check if Calman 2018 is still the same as Calman 2016 which you have? Cool.
 
#37 ·
Yes, exactly. I'll start considering upgrading to Calman 2018 if/when I'll get a brand new shiny 2018 LG OLED with Calman HDR Autocal, SDR 3DLUT and all the bells and whistles.

Thank you for these patterns. It's especially nice that you included the Rec.709 in Rec.2020 patterns as I don't believe those exist anywhere else (as manual patterns).

One request - could you add Rec.709 in Rec.2020 patterns to the color saturation sweeps?
Sure. Give us some time, I'll include it in next patterns build along with several combined versions with different timings.

Also we're continuing our work on Calman patterns (no exact time frame yet but I hope soon) and - hold your breath - patterns with voice tips :eek: Actually we do have a 'prototype' but voice quality is not quite good enough for now.
 
#38 ·
Updated and extended version of HDR-10 Calibration Patterns (HCFR)

Donwload link: HDR10 calibration patterns

Current version includes patterns for HCFR 3.5.1.4 HDR-10 calibration workflow:
- grayscale: 5%, 10%, LG OLED 2016 and 2017 steps
- primaries and secondaries: Rec2020, P3 in Rec2020, Rec709 in Rec2020
- near black 0-20%, 2% step
- near white 96-100%, 2% step
- color saturation 10% and 25% steps: Rec2020, P3 in Rec2020, Rec.709 in Rec.2020
- color checker: GCD Classic, MCD Classic, Calman Classic, Calman SG Skintones, Calman SG, Pantone Skintones

Patterns characteristics:
- MaxCLL: 1000 cd/m2
- MaxFALL: 400 cd/m2
- pattern size: 10%.
- duration: 2min (text label stays on screen for 10 seconds)
- combined version: 10s, 15s, 20s +3s black screen 'cooldown'
 
#39 · (Edited)

Updated HDR-10 Calibration Patterns for CalMAN!

Donwload link: HDR10 calibration patterns



Current version includes patterns for HCFR 3.5.1.4 (see previous post for patterns lists) and CalMAN HDR10 Calibration Workflow:
- White Balance Module (2 points: 30% and 80%)
- Grayscale: 20 steps, 10 steps, LG OLED 2016 and 2017
- CMS Adjust BT.2020
- ColorChecker: Classic, SG, Video and full combined version
- BT.2020 Saturation Sweeps: 20% and 10% steps
- P3 Saturation Sweeps inside BT.2020 20% steps
- BT.709 Saturation Sweeps inside BT.2020 20% steps

Patterns characteristics:
- MaxCLL: 1000 cd/m2
- MaxFALL: 400 cd/m2
- pattern size: 10%
- duration: 2min (text label stays on screen for 10 seconds)
- combined version: 10s, 15s, 20s +3s black screen cooldown
 
#53 ·
- combined version: 10s, 15s, 20s +3s black screen cooldown
These combined patterns are great.

One request - could you make combined versions of shorter duration such as 3s?

To me the ideal would be combined files with hybrid durations that are 10s for dark patterns of say 20% or darker on the gray scale and then 3s per pattern after that. Ted's SDR patterns have 2,6, and 10 second versions but I find that none of them are ideal because what's really needed is 10s for the darkest patterns and then 2s / 3s for all others.

Anyway, if you are able to create some shorter duration combined files for the next version, that would be much appreciated.
 
#42 ·
Yes, Luminance is 50%. Default Calman values.

Also, just in case every saturation pattern has a text label with RGB triplet (TV range 16-235) and it can be compared against Calman expectations. For convenience RGB triplets are displayed as 8 bit numbers (it's what both Calman and HCFR are generating and showing to user) but just to be clear - patterns are created and encoded in 10bit mode.
 
#43 ·
Afternoon folks, I have a probably stupid question regarding these HDR patterns.

When using the white and colour clipping patterns for my display (max 430 nits) should I adjust my display until the closest luminance to my display is the last patch visible (in this case 450 nits) and nothing above 450 can be seen?
 
#44 ·
When using the white and colour clipping patterns for my display (max 430 nits) should I adjust my display until the closest luminance to my display is the last patch visible (in this case 450 nits) and nothing above 450 can be seen?
No, you shouldn't intentionally clip details above your display physical capabilities. Your display should use tone mapping algorithm built in by the manufacturer and be able to display details above 430 nits.
 
#45 ·
Hi!

I want to use your patterns to do a 2pt WB calibration and reference measurement of my LG B7 using HCFR. Could you please help me with these two questions?:

1) Which patterns should I use for measuring grayscale: "HCFR\01. Grayscale\02. 5% steps" or "01. Grayscale\04. LG OLED 2017 steps - 1000nit"?

2) Which preference must be used in HCFR for these patterns?

Thanks in advance!
 
#46 ·
Thanks for the work that you've put into this!

Question, with regards to the Grayscale patterns. What is the difference between the 20 step patterns and the 2016 OLED patterns. I have a 65" OLED E6P (2016) and was curious what are the differences. How do they differ? What's the advantage of the specific 2016 OLED patterns over the normal 20 step, with a 2016 OLED?

Thanks
 
#47 ·
Hi guys!

10-20 steps grayscale patterns have 'generic' 5-10% steps ( 5%, 10% ,15%, 20%, ...).
LG OLED TVs (2016, 2017, 2018) are using very specific codes/values for 20 point WB controls and corresponding calibration patterns are significantly different: 7.2%, 21.69%, 29.22%, 36.76%, 40.53%, ...
Check out calibrations notes from LG:
2016 http://www.lg.com/us/support/products/documents/UHDA Calibration Procedure.pdf
2017 http://www.lg.com/us/support/products/documents/Calibration Notes for 2017 LG OLED TVs.pdf
2018 - ? (Don't have a link.)

So if you're planning on adjusting 20pt WB controls on LG OLED then LG-specific patterns are required.
But if you're planning doing 2 points WB adjustment you can use use 'generic' grayscale patterns with 5-10% steps. However, as far as I know, 2pt WB controls are blocked in HDR mode on 2016 models and only 20pt WB is available.
 
#48 ·
Thanks for the detailed explanation. Much appreciated.

I saw that the TV used codes instead of %, and figured that the LG OLED patterns were just to make it more user friendly and understandable. Now, after seeing your explanation, it never occurred to me that it was because the % values are actually different than the standard 10/20 step. Interesting.

And yes, on the 2016's there is only the option of 20pt WB in HDR.

Again, appreciate the response and all of the work put in.
 
#50 ·
Can someone clue me in regarding the 20p white balance code values/steps on LG OLED 2017 models? This suite contains grayscale patterns for LG 2016 and 2017 for their appropriate code values. I have an LG B7 but the code values in my TV's menu for 20 point white balance don't match the video files in the LG 2017 folder, they match the ones in the LG 2016 folder though. What am I missing here?
 
#52 ·
LG OLED 2018 HDR10 calibration patterns: link.



LG OLED 2017 models actually have not 1 but 3 variants of HDR10 codes for different meta variants: 540, 1000, 4000 nits. But neither of these codes (540, 1000 or 4000) are the same as 2016, so it's not clear why values in the menu match 2016 codes. I think it's worth asking in 2017 models calibration thread.

Also, for now I've built only 1000 nit meta variant for 2017 LG OLED but actually can also add other meta variants if there is any demand for them: 540 (not sure why anyone could need that) and 4000.

Is there a way to burn these patterns to a disc similarly to the AVS AVCHD disc (I have a BD burner so I could also do HDMV)?

If not, anyone have any issues with calibrating an LG OLED via USB with these patterns? I know that they trigger HDR on my E6P, but I don't know if going through USB will produce accurate results.
If you're not planning watching content with internal TV USB player I'd advise not to use it for HDR calibration as results could differ from HDMI. In my case readings from internal TV USB player seem to be slightly brighter than from HDMI source. (But not very stable in any case.)

You can try burning patterns as files/directories to a disc but I'm not sure if it'll work. I've tried creating an ISO with DVDFab UHD creator but unfortunately it was almost unusable - no way to choose which file to play.

So for now I think that the most convenient way is to play patterns from a HDMI player over network or from a USB drive.
 
#51 · (Edited)
Is there a way to burn these patterns to a disc similarly to the AVS AVCHD disc (I have a BD burner so I could also do HDMV)?

If not, anyone have any issues with calibrating an LG OLED via USB with these patterns? I know that they trigger HDR on my E6P, but I don't know if going through USB will produce accurate results.
 
Top