This kind of makes sense as the shader only does work for channels it needs to do when you use a GM mask Vs a RGBX mask the GPU will be doing roughly twice the amount of work as in for every 4 pixels GM calculates 6 channels (0:green,1:blue+red,2:green,3:blue+red) whereas RGBX calculates just three channels (0:red,1:blue,2:green,3:none).
That does make some sense, but then why does the custom GMB mask I was trying out also cause slowdown whereas the RGB mask does not? Consider the following:
GMB: 0:green,1:blue+red,2:none RGB: 0:red,1:green,2:blue
Same number of channels calculated, is it not?
1 Like
Hmm yes that is strange - you just added that mask and did nothing else? When you added it did you just replace an existing 3 pixel mask?
If it’s just a plain swap of the mask then I’d assume it’s something to do with memory accesses that the new mask causes issues with Vs the original RGB mask.
Maybe get Nvidias Nsight/Pix and profile it - see if that shows up the issue.
FWIW, the subpixel mask function has the same behavior. Some of the masks really tank performance and it doesn’t seem particularly clear as to why. That is, some of the heavy ones are big matrices, some aren’t, etc.
2 Likes
Good to know it’s not specific to the Sony Megatron - that is really quite odd though but interesting I’ll see if I can grab a few profiles now I know what to look for.
1 Like
Some pictures of Megatron running in on an LG 55OLEDE6P.
Aperture grill 300TVL
Sony Megatron JVC D-Series HDR
Sony Megatron aperture grill 300TVL
Sony Megatron aperture grill 600TVL
CyberLab Slot Mask (HSM Mega Bezel Reflection Shader/CRT-Guest-Advanced)
The whitish stuff is probably being caused by the bloom. I included it because it shows the individual “cells/elements” which make up each individual subpixel a bit more clearly. The Megatron Slot Mask adjustments seem to be a bit more granular with one able to create a slot mask with 3 vertical elements and a single horizontal black element.
With CRT-GUEST-ADVANCE controls, if I increase the slot mask height by 1 it doubles the height so it would be 4 elements high and with the same 2 element tall black line below.
I think being able to match the granularity of Sony Megatron would go a long way towards improving realism.
Sony Megatron JVC D-Series HDR
3 Likes
Right now I’m not really focusing on the white subpixel because I’m not even seeing any pure white subpixel in any of these pictures and HDR is on. That diagram while being a helpful illustration does not appear to accurately represent some of the nuances between the structure of the different subpixels.
For example I’m seeing the red subpixel consistently being wider than the green and blue subpixels. Depending on which TVL is being used, I’m seeing either 1 or 2 additional columns of red elements used.
There seems to be some strange overlapping going on between the right most red element column and the left most green element column.
I’ve seen an instance where the red elements are off but I can count a group of 3 vertical green elements.
Do note that WOLED subpixel structures have evolved over time. In post 2018 panels, the red subpixel got an increase in size of about 80%.
This might very well translate to masks looking very different between “vintage” OLED TVs and more modern ones.
I have a feeling that the older ones might fare better.
I’m hoping that with the gathering of new empirical data there can be some improvements to be made beyond just black & white masks for OLED displays. It could benefit older TVs with the smaller red subpixel more than the newer ones, who knows, but anything would help.
Perhaps there can be a realignment of the mask pattern to match the extra red subpixel column with the Black (X) of the mask pattern then the colours reordered to match? I wonder how things would look if an RGB Mask, without the X were used.
I’m waiting for someone with a bit more understanding of how things work to assist. I am in a position to provide more test images and samples.
On Megatron aperture grill doesn’t look so bad on my OLEDE6P. It’s just that on certain TVLs the ratio of the width of the red to the other subpixels is 3:2:2 and at higher TVLs it’s at 2:1:1.
Hopefully @MajorPainTheCactus, @guest.r, @Nesguy or @hunterk can chime in.
1 Like
Can you get a pic of a solid 50% gray screen (240p test suite has such a thing, IIRC)?
2 Likes
Sure, just give me a couple minutes.
Ok, I got some pics but stability was a bit of a challenge this time around. I’ll upload these then maybe at some point later when I get a chance I’ll try to do it over a little clearer, however, the results are no different from the ones in my above post which are also of elements which vary from white to grey.
I also took these in a completely dark room. It seems as though the ones from before with at least one light on might have come out better as well.
Aperture Grill, 600TVL
Aperture Grill, 300TVL
Slot Mask, 600TVL
Slot Mask, 300TVL
The first shots were taken using Pro mode standard SMCVM settings, while the zoomed in one’s were taken using Macro mode. The washed out look could be due to the camera not auto adjusting well enough or the light reflecting back and forth between the screen and the camera or something like that.
2 Likes
Hi Cyber hope you’re well! Ok so let me get this straight in my head: the issue you’re seeing is that using the Sony Megatron aperture grille 600TVL preset you’re getting two red pixels and 1 green and 1 blue?
So this is a bit odd - have you tried the 2730 preset? Also what TV are the last posts pictures from? What resolution is it? Have you got all the TVs sharpening turned off and are using full chroma i.e 4:4:4?
2 Likes
Hi, thanks @MajorPainTheCactus and congratulations again! Hope all is going well with you and your family!
I repaired my LG 55OLEDE6P so I’ve been running it through the paces and documenting my findings and observations concerning it’s subpixel structure and how mask patterns are displayed on it.
I’m hoping that now that I’m in a position to provide direct feedback on my particular model WOLED display, that hopefully someone like yourself or @guest.r might be able to imorove upon what is currently in existence for WOLED TVs.
I dunno if they are pixels, you might have to tell me. I was thinking that they might be subpixels.
What I’m seeing is exactly as the picture shows, that at 600TVL the red is wider than the blue and green by one column of subpixels (or pixels, please correct me if I’m wrong).
At 300TVL the red is wider by one column of subpixels as well.
Sharpening on my TV is probably at about 10. I’m using full chroma RGB 4:4:4 in SDR. These photos were all taken using HDR though.
Based on my research and observations my 2016 TV’s subpixel structure is not the same as a post 2018 model because LG increased the size of the red subpixel in those models considerably.
I’m using 3840 x 2160 resolution by the way.
I can’t remember off hand but these might be from the 2730 preset.
At higher TVLs it doesn’t look like RGB at all. I have to flip the mask layout to make it appear a little more like RGB but I don’t need 800 or 1000TVL anyway so I’m not focusing on that at the moment.
The anomalies are not really noticeable at any distance other than right up to the screen so overall the image quality is fantastic. It just feels sad that this beautiful TV can’t seem to get the RGB “phosphor” triads right after I’ve focused so much on that effect during its absence.
It’s almost as if I have to discard or ignore one of my favorite aspects of CRT emulation that I had valued for the past few months. That feeling disappears when I’m actually playing and enjoying my games though.
1 Like
What settings are being used to produce these shots?
1 Like
The first of the pairs of pictures were taken using Pro mode ISO 100, WP 5000K, Manual Focus, 1/60 Shutter Speed.
The second zoomed in one’s were taken using Macro Mode, everything default/auto.
Both sets were taken from a few centimeters away from the screen. With the Macro mode ones being closer.
But that only applies to the last set of 6 I posted on request by @hunterk. The previous ones were probably all taken using the Pro mode settings. Since they are of higher quality. You should probably use those instead of the last set. They all show the same things though.
1 Like
oh sorry, I meant what shader settings. That helps too, though
2 Likes
The only settings I changed in MegaTron besides switching the TVL between 300 and 600 and the Mask Type between Aperture Grill and Slot Mask were disabling all deconvergence and setting both the Peak and paperwhite brightness to 590.
Now that I have your attention, I can rerun some of the tests with the settings you would like to see or suggest.
2 Likes
ok and what scale/resolution are you using?
I’m curious what happens with a higher TVL, three pixel columns per triad.
2 Likes
In RetroArch settings Integer Scale On. Integer Scale Overscale On, Aspect Ratio Core Provided.
My resolution is 3840 x 2160.
No problem, you can look out for that hot mess later. I think I have some shots that I took already though.
1000TVL (I think)
800TVL (Will verify)
600TVL
Either 800/1000 Layout flipped (will confirm later which is which)
2 Likes
weird, ok. I think this sheds some light on some things, I have to do “work” now, though. I’ll get back to this later
1 Like
Sorry, I should have been clearer: I meant 50% gray with no effects on it. The goal being to see the actual subpixel layout without any mask effects interfering, since that’s what we need to know to be able to try and make them evenly spaced.
3 Likes
