Are you talking about the black in the meter or more generally theres a difference? More generally there is most definitely a difference as its running on a 1440p screen and Ive tweaked the min max scanlines a bit to emphasize the black between the scanlines on such a small screen.
1 Like
I meant that you can see the pixel structure on the black area of the magic bar in my image but you can’t see this in your image. Maybe this is the reason why the blacks don’t look right since the pixels are showing even in areas that are black or is this suppose to happen?
2 Likes
Yeah although very subtle I can see non black pixels in your image and I do think thats wrong. Ill investigate to see where thats coming from.
1 Like
There are non-black pixels visible inside the bar in both of those examples. Also whatever Scarf did in the comparison darkened the pixels further to the point where I only see them on the brighter of my two displays. But the original full shot has them clearly visible on both.
Check this to make sure you’re able to see the difference between square 1 and the background: http://www.lagom.nl/lcd-test/black.php
If there continue to be significant issues with full black (or even just because it would be cool to have) maybe you could add a full mask “base level” adjustment setting, independent of the scan lines. Similar to how you’d actually see a phosphor mask up-close in a non-pitch-black room. Then if the fake phosphors are always sitting at say 1 or 2 instead of 0 it might be easier to tell when something’s a bit off.
2 Likes
Oh shit, you’re right, I can see the pixels now. I dunno what happened in Photoshop when I put the pictures together. Sorry about that.
1 Like
Ok so I found the problem and is due to the rec.601 optical to electrical transfer function (yes this is hoing to get technical) or OETF for short. So from the rec.601 Wikipedia page: https://en.m.wikipedia.org/wiki/Rec._601
You can see this equation:
So far so good so a while back people were complaining that they werent getting an even grey ramp up and when values got small i.e dark they were seeing ‘bumps’ in the gradient when using non 2.22 gamma’s at least when quite a bit off. Essentially when I looked into this I noticed quite a few bars were falling into the the less than 0.018 equation i e 4.5*Luminance part and so the ‘bump’ was the transition between above 0.018 vs below it.
So I thought why not just remove the below equation and just have use the above 0.018 luminance equation. Perfect ramps now were a thing. However the problem with the above 0.018 equation is that if you have a luminance of 0 i.e black it ever so slightly brightens it to a non zero value. This is what you are seeing.
So what to do?
Reintroduce the proper standard and have black blacks but with a bump in the gradient when using more extreme gammas for extra brightness or keep it as it is and rely on people lowering the gamma until its not really noticeable but you have smooth luminance gradients.
I think reintroducing the full rec.601 gamma standard and having black blacks is the way to go and then offering a non standard option that gives smooth gradients. I could also ditch the standards and just use straight power curves for gamma. 🤷
3 Likes
Would it help to do a wet/dry mix between the proper standard and the no-below method that’s tied to the gamma value? That is, as they get into the extreme range, it’s more no-below but in the normal range, it sticks to the proper standard?
automagic stuff like that isn’t always a good idea, but it seems like it might be worthwhile here.
2 Likes
Maybe keep gradients smooth as default and have the other way be optional. Add a way for users to adjust the full phosphor mask level, and by default tie it to the base black level, so black scan lines are the same color as the mask. The issue is more when non-zero black scan lines are standing out separate from their background. In addition to this, some noise might help, especially if you could make it like the kind you see on real CRTs as opposed to the more plain “film grain” type.
2 Likes
So I think what Im going to do is add a simple inverse gamma power curve to get to linear space i.e pow(luminance, 1/gamma). This will have perfect blacks and a smooth gradient out of the box in exchange for a technically non standard gamma conversion.
Ill keep the options for rec.601 and rec.709 OETFs and add an additional option to bias the break off point i.e the 0.018 values. Annoyingly this value is different between the two standards and so having a direct value is problematic. I might have another think about that particular issue as its more a UI issue.
Comments welcome.
3 Likes
Most of what I do, I do by trial and error so I say try it. Once it’s out in the wild you’ll get the feedback that tells you whether it works well or not.
1 Like
The image was a little washed out, but the camera exaggerated it. I didn’t realise how washed out it was until i used 240p test suite and noticed the raised blacks. I’ve set hdr to 720/200 and gamma to 2.5 which has deeper blacks at the expense of overall brightness.
2 Likes
@barack-ojama @Scarf @Wilch who have a LG CX and everybody else that might be interested I’ve just created a pull request with V5.5 of the Sony Megatron in it.
This reintroduces all the standard OETF and EOTF functions and has perfect blacks and smooth gradients. The trick was to add a parameter for the gamma ‘cut off’ point and allow the user to define where this should be and set it to be a lot lower so only black values are included in the less than. I’ve written in the parameters what these ‘gamma cut off’ values should technically be according to the standard but I’ve set them all to 0.001 so that only blacks go into the ‘less than’ equation.
Because RA shader parameters UI only shows numbers to two decimal points I just multiplied the values by 1000 so say 0.018 is a value of 18 in the UI. If anybody knows how to get more precision shown in the UI I’m all ears.
One thing I’ve noticed on my BenQ EW3270 is that ‘mask accurate’ now works in HDR with the proper gamma and luminance values. It doesn’t warp the colours on that display - I might recheck my others to see what happens there too.
Please do try this again and play around with all the new parameters I’ve added. Let me know how you get on. Oh and I’ve broken out the white temperature into two values one for D65 and one for D93 and not to be an offset anymore but a direct value.
5 Likes
Looks much much better - hopefully you can get even better results with my latest V5.5 - play around with the CRT ‘inverse gamma cut off’ with low values say less than 10 in the grey ramp to get smooth gradients + black blacks and brightness - also try mask accurate.
1 Like
I haven’t found a way to do this either. In the Mega Bezel most of my 0-1 values I set as 0-100 with 1 or higher as the step, so the UI is easier to use, and easier on the eyes.
If later, I find it needs a bit more precision I can always change the step to a smaller value since there are 2 decimal places visible in the UI.
3 Likes
Even if the UI doesnt do what we want at least our work arounds are the same! Thats exactly how I solved this particular issue too - I represented 0.001-0.1 as 1-100 with steps of 1. Not ideal but workable.
3 Likes
Thanks for the link @DevonCM - Id forgotten about this website! Ive just been going through all the tests with mobile phone display and it shows just how absolutely amazing mobile phone screens are - at least the AMOLED one in my OnePlus 8 Pro. When the brightness is shoved up it passes most of the tests with flying colours compared to other screens Ive tested them with in the past.
1 Like
I just came across this prototype “DPT” microLED technology. https://www.porotech.com/technology/
It looks like a really good idea, I hope it works out. I’d love to be rid of subpixels.
4 Likes
Looks very interesting thanks for the link. Id be interested in how it copes with text as currently with sub pixels youre getting triple the horizontal resolution to render all those curves - possibly this could be better if each micro led does gradients but if not then I worry itll have too much aliasing. I guess we’ll find out! Any displays planned?
2 Likes
It sounds to me like it’s intended for applications that are currently covered by direct-view LEDs, rather than TVs/monitors (at least right now). This suggests they’re currently pretty large.
And yeah, I guess it depends on the size of the microLEDs and whether the application is subpixel-aware. The RGBx masks should be fine, but the magenta-green masks would not be. That is, if they’re the size of a current subpixel, it’ll represent a pretty massive boost in resolution, but I suspect they’re meant to be the size of one current pixel, which breaks all of the subpixel tricks.
3 Likes
So my account got nuked for no reason. I presume it had to do with my reply to hunterk being detected as spam since I tried to delete it once while it was getting approved to correct something and it was taking too long to get approved at first. I know @MajorPainTheCactus liked my first post that actually got approved
Anyway, I guess I’ll try to reiterate what I posted: I’ve known about Porotech for a while now. Karl Guttag wrote about them recently and said glowing things about them.
Somewhat more on topic: Saw this recent low TVL 3.7 inch CRT PVM and I’ve never seen the Super Mario All-Stars logo look so smooth
I’m convinced more than ever that once we get to ~12000PPI displays we won’t even need shaders anymore
It is based on MicroLED because of their fast response time which is in the nanoseconds to begin with.MicroLEDs are so fast that they can change colors virtually instantly. It’s DLP Color Wheel on steroids. All the commonplace subpixel tricks in shaders would work but at a reduced resolution.
To utilize the maximum resolution you would have to write a shader tailored to the DPT technology obviously. But by the point DPT becomes mainstream there might not even be a need for CRT shaders anymore(besides maybe Composite/RF simulation but that’s different). Once we hit +10000 PPI on displays the pixels are so small that they basically reconstruct the original image without any noticeable jaggies or artifacts.
Now let’s hope my account doesn’t get nuked again for no reason.
1 Like