AVS Forum banner

281 - 300 of 1220 Posts

·
Premium Member
Joined
·
9,832 Posts
One last thing: I've finally been able to do more tests with the latest 3.1.0 and I agree that more than 1400 frame rolling average isn't necessary. Very stable, with just a bit of adjustments.

I went all the way down from 100,000 and even 2160 (90 secs) is still too much and unnecessary, at least with the content I watched.

So here are my updated settings for V3.1.0 (real peak brightness 100nits):

Chapter Flo
Real/minimal target: 100
Maximal Target Nits: 4000
Rolling avg: 1440
Dynamic Target Nits Tuning 100
Chapter detection Fall change in %: 100
 

·
Registered
Joined
·
787 Posts
One last thing: I've finally been able to do more tests with the latest 3.1.0 and I agree that more than 1400 frame rolling average isn't necessary. Very stable, with just a bit of adjustments.

I went all the way down from 100,000 and even 2160 (90 secs) is still too much and unnecessary, at least with the content I watched.

So here are my updated settings for V3.1.0 (real peak brightness 100nits):

Chapter Flo
Real/minimal target: 100
Maximal Target Nits: 4000
Rolling avg: 1440
Dynamic Target Nits Tuning 100
Chapter detection Fall change in %: 100

Thanks for your comments and sharing. Best way like you said would be to compare few titles and switch between modified measurement files.



Did you play with Dynamic Target Nits Tuning values? I see yours is set to 100%. :)
 

·
Premium Member
Joined
·
9,832 Posts
Thanks for your comments and sharing. Best way like you said would be to compare few titles and switch between modified measurement files.



Did you play with Dynamic Target Nits Tuning values? I see yours is set to 100%. :)
It's not 100%, it's 100 nits :)

Yes I tried 50nits and 75nits, but it was too bright for me. I think 50-60nits as suggested by Soulnight should be fine for you, but as I say if it's too bright you can raise it.

I have played a bit with the %fall value but 100% is fine with 3.1.0.

You don't need to switch measurements files, you just need to use Soulnight's utility with a single file, try different values, and play the title after each calculation. Then try another title, and see what you like best.

Or put a few titles in a sub-folder, and select that sub-folder as the path in the utility. That way it will remeasure all your test files with the same value. That's what I do.

Each title only takes 5-15 seconds provided the measurements are already taken. The measurements only need to be done once. The optimizations can be done as many times as you want, and they are very quick.
 

·
Registered
Joined
·
787 Posts
It's not 100%, it's 100 nits :)

Yes I tried 50nits and 75nits, but it was too bright for me. I think 50-60nits as suggested by Soulnight should be fine for you, but as I say if it's too bright you can raise it.

I have played a bit with the %fall value but 100% is fine with 3.1.0.

You don't need to switch measurements files, you just need to use Soulnight's utility with a single file, try different values, and play the title after each calculation. Then try another title, and see what you like best.

Or put a few titles in a sub-folder, and select that sub-folder as the path in the utility. That way it will remeasure all your test files with the same value. That's what I do.

Each title only takes 5-15 seconds provided the measurements are already taken. The measurements only need to be done once. The optimizations can be done as many times as you want, and they are very quick.

My bad :eek: :D Thanks for your very useful comments and suggestions ;)
 

·
Registered
Joined
·
204 Posts
By the way, I use 4000nits max, but I have 100nits peak brightness, not 50nits. I'm only doing this to see if I find anything brighter than the Meg, to see if 4000 is too much or not.

2500 is probably a more realistic max.
Just as FYI - I checked my collection of aprox 100 UHD files and only The Meg required more than 1500. So right now I am leaving 1500 for all titles but The Meg, and I do a quick check with any new titles I rip moving forward.
 

·
Premium Member
Joined
·
9,832 Posts
Just as FYI - I checked my collection of aprox 100 UHD files and only The Meg required more than 1500. So right now I am leaving 1500 for all titles but The Meg, and I do a quick check with any new titles I rip moving forward.
Yes, the Meg is by far the brightest title at the moment (I have around 300 UHD Blurays), but there is no need to lower the max for other titles. A title like Pacific Rim goes above 1300nits and never looks dim, so with 100nits of actual peak brightness or more, I see no reason to limit the target to 1500nits. I might use a 2500 once I find a title brighter than The Meg, as I suspect it might be my limit.

If I had only 50nits like Soulnight, I would most certainly use 1500nits for all my titles, including the Meg.

The point is, you don’t want the picture to become too dim just to resolve highlights. But if it’s not too dim, it actually looks better with a 2000nits target for these scenes, at least here (dedicated room, 0 ambient light, 100 actual nits). Far more detail in the sky if I don’t limit it.
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #289 (Edited)
madmeasuredynamicclipping Tool V3.2.1

And now a nice update so you have something to play with on Sunday:

V3.2.1

Download:

(re-released with 2 bux fixed in V3.2.1)
http://projectiondream.com/wp-content/uploads/2019/01/madMeasureDynamicClippingV3.2.1.zip

NEW:

1) Added new window/form when optimizing a movie to see which movie is CURRENTLY getting processed.
(THE main form behind is displaying the last finished movie)

2) Added *.Webm format to the file which can be measured (Note, we didn't test it yet since we don't have such files)

3) Changed "Choose Path(s) from 5 to 10 paths selectable

4) Fixed bug discovered by @sonichart in combination with the Masciola pattern (or VERY special kind of movie).
-->case 1: a movie has no highlights whatsoever over 100nits during the full movie
-->case 2: a movie is completely black from start to finish :D

5) Introduced "Auto-start" option. If checked, next time you execute the exe, the optimization will/should start directly scanning the paths saved in "choose patrhs" with the latest settings used. ;)

6) Added option to select "Minimal chapter duration". (was hardcoded to 240 frames in V3.1.0)

7) Added option to merge chapter if the relative change in % of "avgFALL" between 2 consecutive chapters is LESS than the X% value selected. Default value 50%. (was hardcoded to 50% in V3.1.0).
-->This may render the "Scene merge" (previously called "Chapter detection") redundant. You may try to put "scene merge" to 0% and only work with "chapter merge".

Difference between the two:
- "scene merge" only compare the FALL value for last frame of a scene and the first frame of the next scene.
- "chapter merge" compare the AVGFALL over a whole chapter/scene and compared it to the avgFALL value of the next chapter/scene

8) Changed slightly the centered rolling avg algo. Before if you were chosing 10s rolling avg (240 frames in 24p) and the chapter was 15s long, the first frame was calculated with the next 10s/2 future frames avg. Next frame would have been calculated with the previous frame + the 10/2 future frames and so on.
The issue is that you were doing the avg effectively on less than the chosen number of frames at the start or the stop of a chapter.

Now in V3.2.0, we would apply (in this case) for all the frames in the first 5s, the avg of the 10 first seconds of the chapter.
This ensures more stability and ensures that you do not build an average on less frames than what you had selected in the GUI. (except if the chapter is shorter than the the rolling avg time)

A a consequence, we modified the defaut value for rolling avg in frames to 480 (20s) frames instead of 1440 frames.
(with 1440 frames (1min), in V3.1.0, for a long chapter, the first frame target nits would actually only have been calculated with the average of the first 30s and not 1min)

9) Removed "Flo Scene". But this can still be activated if you put 0 as settings to "scene merge", "minimum chapter time" and "chapter merge".

10) Added to details file, the "number of chapter" generated for the "dynamic target nits" logic. With this you can see how the change in the settings impact the chapter split

11) the dynamic target nits algo now always refers to the REAL original peak even if the dynamic clipping option is checked. ;)

12) Solved 2 additional bugs reported by @Dexter Kane

-->Please use the button "Restore original measurement file" BEFORE starting a new optimization. (Detail file structure has changed).

Here the new GUI.




Enjoy,

Anna&Flo :)
 

Attachments

·
Registered
Joined
·
173 Posts
@Soulnight I'm having some problems with the new version, I get an unhanded exception (index was outside the bounds of the array) on some measurement files. It seems to create the new measurement file but not the details file.

Also the first section of the details file is the same for each movie, at least when using a path containing multiple movies. here's an example

Code:
madMeasureHDR output (9/01/2019 6:31:47 PM)
Metadata
Mastering display luminance: 0.002/1000
MaxCLL: 0
MaxFALL: 0 nits

Measurements
Frames: 75767
MaxCLL 100%: 893
99.9%: 593
MaxFALL: 232
AvgFALL: 9
AvgFMLL: 135 nits
----------------------------------------------
Header
Version: 6
Header size: 10
Number of scenes: 3963
Number of frames: 149006
Complete flag: 1
MaxCLL: 1404
MaxFALL: 307
AvgFALL: 7

The first section before header is the same for every details file the rest of the file seems to be correct.
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #292 (Edited)
@Soulnight I'm having some problems with the new version, I get an unhanded exception (index was outside the bounds of the array) on some measurement files. It seems to create the new measurement file but not the details file.

Also the first section of the details file is the same for each movie, at least when using a path containing multiple movies. here's an example
Ok. We changed a bit the logic to add new compatible video format.

I think it could comes from there.

I have removed the download link.
Will be added back when the bug is fixed


Ps: you did use the "restore original measurement file" before pressing "start", correct?
;)

Edit: could you please upload one of the problematic measurement file? Thx.
 

·
Registered
Joined
·
173 Posts
Ok. We changed a bit the logic to add new compatible video format.

I think it could comes from there.

I have removed the download link.
Will be added back when the bug is fixed


Ps: you did use the "restore original measurement file" before pressing "start", correct?
;)
Yeah I restored the original files before I ran it the first time
 

·
Registered
Joined
·
173 Posts
Ok. We changed a bit the logic to add new compatible video format.

I think it could comes from there.

I have removed the download link.
Will be added back when the bug is fixed


Ps: you did use the "restore original measurement file" before pressing "start", correct?
;)

Edit: could you please upload one of the problematic measurement file? Thx.
Here you go, it was only two that gave me problems

https://owncloud.aquinasnet.com/index.php/s/SoTVXLvtvmKmxrX
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #296

·
Registered
Joined
·
1,479 Posts
Discussion Starter #297 (Edited)
madmeasuredynamicclipping Tool V3.2.1

So,

V3.2.1 with 2 bug fixed. Thanks @Dexter Kane

As always, please do a "restore all original file" before pressing "Start" with a new version of our tool.

Bug Fixed:
1) Serious:
In a very special case with V3.2.0, the new chapter logic could crash. Please re-optimize your files with V3.2.1

2) Harmless:
We always save & keep the "madmeasureHDR" output in the header of our detail file. However we can nly keep it if you did generate the measurement files with our tool.
The bug was if you did not measure some of them with our tool, but some other with it, the header from the one you had measured with our tool was copied to the ones you did not measure with our tool (if they came afterwards)


Download
http://projectiondream.com/wp-content/uploads/2019/01/madMeasureDynamicClippingV3.2.1.zip
 

·
Registered
Joined
·
1,209 Posts
Thank you!

I was thinking if it wouldn't it be more logical to have the values in seconds instead of frames?
But does your tool know the framerate of the analyzed content?

Also a very small request: can you tool remember that I ticked/unticked "Apply dynamic clipping" and "Apply target nits selection" the next time I launch it?
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #299
Thank you!

I was thinking if it wouldn't it be more logical to have the values in seconds instead of frames?
But does your tool know the framerate of the analyzed content?

Also a very small request: can you tool remember that I ticked/unticked "Apply dynamic clipping" and "Apply target nits selection" the next time I launch it?
1)
Yep, our tool cannot know the frame length in seconds since everything in the measurement file is written in frames and not in seconds.

We could maybe grab some information directly out of the mkv but is that worth it?

2) Sure this could be added later.
But why would someone want to always keep uncheck one of those? ;)
 

·
Premium Member
Joined
·
9,832 Posts
1)
Yep, our tool cannot know the frame length in seconds since everything in the measurement file is written in frames and not in seconds.

We could maybe grab some information directly out of the mkv but is that worth it?
The main advantage would be that the algo works the same for all files, irrespective of their frame rate.

For movies, it doesn't matter as there is only one 60p movie (Billy Lynn) and it's dreadful. But who knows, there might be a good one released at some point (the new Avatars?)

For demo files, it's more relevant as many of them are 60p.
 
281 - 300 of 1220 Posts
Top