I couldn’t find references for your XYZ_to_sRGB coefficients. AFAIK the transformation matrix is:
3.2404542, -1.5371385, -0.4985314,
-0.9692660, 1.8760108, 0.0415560,
0.0556434, -0.2040259, 1.0572252
2 Likes
You must consider that GLSL/SLANG matrices have to be transposed. Rows become columns or vice versa.
2 Likes
Ah yes, I recall having to do that before. Didn’t notice it was transposed. 
2 Likes
Is the latest update available for download yet?
I noticed that disabling hotspot fix and lowering contrast by .05 seemed to give better results but I’m maxing out my display backlight so it’s kind of a unique situation; not sure how applicable my results are to the typical use-case. Personally, I think the backlight should always be maxed out when attempting any kind of CRT emulation, but not everyone agrees.
1 Like
Yeah, it’s in his GitHub.
1 Like
Yes, I updated yesterday you can check the commits. The problem with hotspot fix was the transition, it wasn’t smooth so sometimes a line could be seen. I need to design something better but all in all the whole reason was to fix something that shouldn’t be there to start with, so I need to study the case further. I think that for some reason with my latest changes the situation is better but have to do more tests.
For the LED backlight I don’t know, you have done more research on that. I just wonder CRTs were like 100nits, and current LDR LEDs are between 200 and 300 nits. What am I missing? The important thing to note is that the signal wasn’t normalized, all pure white created was on the CRT side.
1 Like
oops, didn’t realize I had two copies of grade and I replaced the one that my preset wasn’t pointing to.
Yeah 300 nits is about right for most LCDs unless they’re really old. CRTs were 100 nits with masks and scanlines, though- once you add that stuff to an LCD it causes a substantial drop in luminance
Trying to correct this in software tends to introduce clipping but if you just crank up the backlight then you can increase the luminance while leaving the color values unaltered… That’s about all there is to it.
1 Like
SMPTE-C seems way too bright and washed out to me… what am I doing wrong?
POW and sRGB gamma seem fine. Also, should LCD gamma be 2.00 or 2.20?
You need to refactor your current settings. I don’t know how your display is calibrated but if it approximates a gamma curve of 2.2 it should be fine, otherwise you might have a very lifted knee.
TV settings wise, brightness at default 50, contrast to 100, and backlight 10 for 120 cd/m2 (nits), raise for higher nits. Then there’s a gamma curve option, when calibrating I found a -1 or -2 is appropiate. Set to -2 if raising back light. With this LCD gamma should be 2.2.
1 Like
Does the SMPTE-C setting need a wide color gamut monitor to look right? I seem to get clipping or crushed blacks no matter what I do.
Just enabling POW gamma seems to result in the most accurate black level and contrast. Even sRGB looks a bit bright, but this display is pretty wacky.
1 Like
Maybe it’s the display, is it calibrated?
In any case I will run some tests between today and tomorrow but the function transform seems legit as it’s extracted from this page in wiki. The NTSC (FCC) uses a power gamma of 2.2, so you can try that as well.
There’s also the LCD gamma setting to adjust to your display gamma, or play with grade contrast or black level. I mean what looks best to you, I’m still trying to color manage retroarch without success, and things might change or not, but this is a personal quest not something I will translate to the shaders.
2 Likes
Oh it’s definitely the display, no doubt about that. Greens are badly oversaturated post-calibration. Yeah, I’ll probably just need to keep tinkering and will probably never be fully satisfied until I can upgrade to a wide color gamut monitor. 
I’ve got a stimulus check burning a hole in my pocket, maybe I should go buy a new display and stimulate the economy, lol. Thanks for the tips!
3 Likes
Go buy a display, lol. Your constantly complaining about your current ones so might as well as long as it doesn’t hurt you financially, 
.
1 Like
I’m about to buy a new colorimeter i1 DisplayPro Plus, so I can calibrate my TV as well. I think it’s a better investment over a TV unless the TV is crap… Then I can give better answers on the whole workflow.
2 Likes
Yeah I’ve just had horrible luck with these things . I was pretty happy with the Plasma, though. Biggest drawback was the temporal dithering which messed with masks, and just a general lack of brightness. It’s living in a storage unit now due to my current chaotic living situation. 
I get bad buyer’s anxiety with these things but, screw it. I can’t live with these compromises anymore lol. Time to spend some $$$
1 Like
Just find all the reviews, lol.
I understand your struggle with buying things, I’m super anxious when I start spending more then $40 on something.
1 Like
Hey I gave myself some leisure time to check everything and I can confirm, SMPTE-C function transform is not for everything. Testing with Super Mario Bros (and other games) I found that POW at 2.2 is better (like the FCC standard), 16 bit for the most part I find SMPTE-C is they way to go, and everything else sRGB (with a few exception on PSX games)
2 Likes
I think I pretty much got SMPTE-C looking how I want, I just had to lower LCD gamma to 1.80 and black level -20. Looks really good to me with my backlight @ 75%; plenty of brightness to spare for a mask effect.
SOTN might actually need sRGB, but oh well.
3 Likes
Looks great to me, good tonality, now I can see the bricks in SOTN.
1 Like