Page 1 of 1

A curious observation (contrast, RGB gain) on my ASUS

Posted: 12 Oct 2017, 14:11
by flexy123
I have an old(er) Asus VE248h where I just made a curious observation.

In the OSD (or with 3rd party tools using DDC/CI commands to control the monitor), you have levels for R, B, G (gain), ranging from 0-100, right.

However, these 100 levels for setting RGB gain are not actually applied, but in reality only 23 steps across this range. (I can check this visually by increasing G gain in the OSD, it always goes in little steps approx. 4 levels apart but nothing happening in between). (For example, I see a change at 28, then 32...and so on).

Now, NORMALLY this wouldn't be an issue.

The issue comes into play when you are adjusting CONTRAST, which I understand is also gain, but then for RGB combined.
Contrast however does full 100 steps.

As a result of this, there is an inconsistency in colour accuracy when you adjust contrast. Say, you gradually increase contrast and then observe how the display shifts repeatedly from having a slight pink hue, to having a slight green hue, from one value to the next and repeating. (Such as contrast 20 slightly greenish, 21-22 normal or pinkish, 23 again greenish etc.)

Say you calibrate the display at one contrast setting "ok", but then you change contrast to one of these contrast values, and your calibration is off again. I have NEVER seen such happening with a monitor! (I can now do a quick and dirty visual test with just slowly sliding contrast and see these colour shifts going on all across the range). (I am quite OCD when it comes to colour accuracy)

Has anyone ever observed something like this? Is it normal that some monitors although apparently providing 100 steps of RGB gain actually only provide 25 steps? (I am also not sure whether this affects also R and B gain, but with G gain it's just most obvious from looking.)

(Fortunately, this strange behaviour doesn't affect brightness values....and it is ONLY an issue if one happens to constantly change their monitor contrast, which NORMALLY doesn't happen anyway. I was just stumped that with this monitor I never got calibration (blacks, whites etc.) right because I change these values around all the time (eg. at night)...and now I know the reason for this inconsistency). Just think it's really odd.

Re: A curious observation (contrast, RGB gain) on my ASUS

Posted: 12 Oct 2017, 16:13
by Chief Blur Buster
This is normal until you use 10-bit or 12-bit LUTs.

Adjusting digital contrast can cause minor hue shifts during low-precision lookup tables, or low precision math, especially 6-bit.

Mathematically, imagine trying to divide RGB(63,62,59) by two if you're trying to halve contrast, you're ending up getting RGB(32,31,30) which isn't an exact math for each color component. That means a hue shift.

Now if you have a 10-bit LUT, then you're working with RGB(1023, 1000, 975) and dividing that by two results in RGB(512,500,488) ... that's a smaller imprecision than during a 6-bit LUT (RGB 0..63 range) than 10-bit LUT (RGB 0..1023 range).

Hue shifts are unfortunately rounding errors. It's normal during picture adjustments on a low-precision LUT. :(

It doesn't affect Brightness because it's simply integer math. Adding +2 to RGB(32,31,30) is simply RGB(34,33,32). So you don't have hue shifting during a linear brightness shift because it's 100% integer math. Unfortunately, Contrast adjustments necessarily has integer math rounding error = major hue shifting errors during low-precision.

Ideally, monitors should internally process in full 10-bit or 12-bit colorspace BEFORE downconverting to 8-bit FRC (for 6-bit panels) with proper temporal dithering to reduce hue shifts.