| Forums | Register | Polls | Search | Statistics |
Display Calibration Light Illusion Forums / Display Calibration /  
 

User Error - Patch Scale

 
 
Page  Page 2 of 3:  « Previous  1  2  3  Next »

Author Steve

INF
Male
#16 | Posted: 5 Feb 2021 16:18 
liberator72
Yes, I see that now - unfortunately the required info was missing...

Steve
Steve Shaw
Mob Boss at Light Illusion

Author Steve

INF
Male
#17 | Posted: 5 Feb 2021 16:24 | Edited by: Steve 
Here is an Excel that may help those that struggle with understanding range conversion.
(It can be used to convert % into integer range (and vice-versa) too - just set the Input/Output values as required)

Steve

Patch Scale Conversion.zip Attached file:
Patch Scale Conversion

 
Steve Shaw
Mob Boss at Light Illusion

Author liberator72
ZRO
Male
#18 | Posted: 5 Feb 2021 16:24 
ebr9999
This is what I tried to explain this morning in our conversation. Unfortunately, as I was working at the time I probably didn't explain it as clearly as I maybe could have.

The bottom line is this. Once you understand that when using a custom created patch sequence it matches that of the Patch Scale setting in ColourSpace, everything works exactly as it did before, with the added benefit of no rounding errors displayed in the GUI of the software (due to the 0-255 -> 16-235 target mismatch).

Author liberator72
ZRO
Male
#19 | Posted: 5 Feb 2021 16:27 
Steve:
Yes, I see that now - unfortunately the required info was missing...

No worries Steve.

Just trying to help explain. As you know I was dealing with all this last week when the first build with these changes was released, and I too created an Excel for this purpose.

While it may seem confusing at first, once understood it is very simple and has more benefits than pitfalls (just my opinion).

Still some kinks to work out, as we have discussed, but I am sure they will be resolved in due course.

Author ebr9999
ZRO
#20 | Posted: 7 Feb 2021 11:16 | Edited by: ebr9999 
So, Steve, from now on CS is accurate also for levels different than PC? (indeed not the case for LS)
Somehow a good piece of news. Honestly, I don't understand why you have not taken into account our (as users) legacy. I see that in measurements files you have kept 0-1 stimulus scaling, representing integer 0-255 RGB stimulus. So nothing has changed, hence you need scaling and rounding according to the patch scale (you have added as metadata). Anyhow it is only a rhetorical question from a lazy user

Now, as beta tester, a first question: how can we test that? AFAIK there is not any log displaying what sent to HW and used for your calculation.

What I have noted, using your converter (it requires the use of ROUND for being perfect), is the following:
Patch Scale Levels
16-235 ... 213 214 214 215 ...
↓ ↓
0-255 ... 229 231 231 232 ...
I.E, nothing new; level 230 does not exist in that conversion. But if read it with 0-255 scale and then switch to 16-235? Have a look:

Second question: I know that one of those 214 levels should not exist, but why they have different same RGB8 and different targets and dE? That makes me suspicious the right rounded value has not been used.

You can find the .bcs as annexed

EDIT: I have been convinced the second question does not have sense. One of those 214 (the one inserted by me with patch scale 0-255) does not match exactly level 214, as it should. From that the different (incorrect) targets. Looking at data in the measurement file (bcs) I am everything works as intended.

PS:
What I mean for accurate: the stimulus stored in measurement files and sent to the HW (pattern generator) allows for perfect matching between what output by the HW and input of CS internal calculation like dE, 3DLUT generation. I have written "allows" as it is up to the calibrator to set HW and display accordingly, also keeping in mind that CS resolution is currently limited to 8 bits.

230-231 (new Release) = 16-2.zip Attached file:
bcs file

 

Author Steve

INF
Male
#21 | Posted: 7 Feb 2021 11:22 
I actually don't understand you questions... Sorry.
But hopefully, all explained on the website: https://www.lightillusion.com/patch_scale.html

Steve
Steve Shaw
Mob Boss at Light Illusion

Author ebr9999
ZRO
#22 | Posted: 7 Feb 2021 11:37 
1) How can I check CS is accurate (see my definition of accurate)
2) How is that level 214 of 16-235 levels appears twice with the same RGB8, but with different targets in my snapshot?

Author Steve

INF
Male
#23 | Posted: 7 Feb 2021 12:56 | Edited by: Steve 
ebr9999:
1) How can I check CS is accurate (see my definition of accurate)

I guess same as with any other system?
Use a HDMI analyser?

ebr9999:
2) How is that level 214 of 16-235 levels appears twice with the same RGB8, but with different targets in my snapshot?

Because the patch data holds two patches... One will not magically disappear.
The rounding will alter the triplet value, as it is integer, but the measured data remains the same.
As it says in the link provided, Patch Scaling must be the set same as when the profile was made, or there will be rounding - obviously.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author liberator72
ZRO
Male
#24 | Posted: 7 Feb 2021 13:49 
ebr9999:
But if read it with 0-255 scale and then switch to 16-235?

As I have tried to explain before, any patch set you use must match the Patch Scale setting you use. If you take one set of measurements and then change the Patch Scale, there will obviously be rounding errors.

This is why I said before that you must pre-scale any patch sequence based on the Patch Scale setting you wish to use (or re-scale any previously created patch sequence).

If you look at BCS measurement file now, you will also see that the saved data is also correctly scaled depending on what you set as Patch Scale.

For instance, with Patch Scale 0-255, triplet value 163 is normalised to 0.639215686 (0-1 range). But when you scale 163 to Video Levels it becomes triplet value 156. and when you set Patch Scale 16-235, it is correctly recorded as a normalised 0.639269406 (0-1 range).

So the measurement data is correctly scaled and recorded as such. There are differences between 0.639215686 and 0.639269406 because there are differences in the range of 0-255 and 16-235.

This also means if you measure data with one Patch Scale setting and then change that setting afterwards, there is simply no way to avoid these rounding errors. You must match the patch sequence to the Patch Scale setting, and not change the patch scale setting afterwards. Anything else will be wrong.

Author Steve

INF
Male
#25 | Posted: 7 Feb 2021 15:00 
I think there is a bigger issue being overlooked here.

If you have a displays set to expect a limited range (16-235) input signal, the signal path will be processed differently to when the display is set for a Full range (0-255) input signal.

The processing has to be different, as the screen's black and white levels have to mapped according to the input signal.
If it wasn't, black would be a lifted grey, and white would be dull, with a 16-235 input vs. 0-255...

You cannot have one set of profile data work in both configurations.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author ebr9999
ZRO
#26 | Posted: 7 Feb 2021 15:04 
Steve:
I guess same as with any other system?
Use a HDMI analyser?

Sorry, a stupid question from me: I can see that in measurement logs

Author Steve

INF
Male
#27 | Posted: 7 Feb 2021 15:06 
ebr9999
No, an HDMI signal analyser is a totally separate device...
https://www.lightillusion.com/murideo_six-a.html

Steve
Steve Shaw
Mob Boss at Light Illusion

Author ebr9999
ZRO
#28 | Posted: 7 Feb 2021 15:37 
Steve:
You cannot have one set of profile data work in both configurations.

I understand that, but let me say that as beta tester, I try also digging more.
Of course, you only know your sw, so I can only try to spot something I cannot prove (otherwise I would report it as a bug).
I have inserted an RGB8 stimulus in bcs (a trick) whose corresponding RGB8 in 16-235 is the same as a valid one. I am not questioning as they are two. I am only spotting they have different targets, despite RGB8 are the same. But it is up to you to tell, don't worry, all related calculations are based on RGB8 as intended.

Author ebr9999
ZRO
#29 | Posted: 7 Feb 2021 15:40 
Steve:
No, an HDMI signal analyser is a totally separate device...

I was asking about levels sent to HW (RPG), not generated by HW.

Author Steve

INF
Male
#30 | Posted: 7 Feb 2021 15:47 
No, when you change the Patch Scale range, only the triplet values are rescaled, not the measured data.

So, with the two values in your test profile, as they are on the rounding boundary, one of the triplet values changes to be the same as the other, while the measure data remains as was actually measured.

That is exactly as expected.

As the user guides state, the Patch Scale for any recalled profile MUST match the Patch Scale used when the profile was made, or the triplet values will be rescaled.

It is that simple.
There is no bug, and no unexpected behaviour.

Steve
Steve Shaw
Mob Boss at Light Illusion

Page  Page 2 of 3:  « Previous  1  2  3  Next » 
Display Calibration Light Illusion Forums / Display Calibration /
 User Error - Patch Scale

This topic is closed. New replies are not allowed.

 

 
 
Online now: Guests - 1
Members - 0
Max. ever online: 264 [6 Jun 2020 12:02]
Guests - 264 / Members - 0