@Dennis1 What software / calibration tool are you using with your i1Pro 2 to do the measurements in the procedure quoted below?
1 Like
This is really interesting to hear! So all the console manufacturers just ignored the standards? Actually I dont know that as I havent tested but certainly on the SNES if the bsnes-accuracy core is correct then nintendo didnt.
Hi @Azurfel Ive finally caught up! So I haven’t tried out lilium hdr tests and am not sure how they work but Im very interested in any bugs I need to squash. So Im guessing youre talking about all the black dots inside the rec. 709 triangle? Purely guesfing but that kind of looks like floating point precision issues as it kind of looks like its repeating.
What are you testing here? The Reshade or RA version of Megatron or is this just the HDR support of RA? Are you testing a particular test pattern such as those in the 240p test suite? Sorry if these are stupid questions I need to get this test suite and have a play.
1 Like
Great stuff one more question then: have you made all the scanline settings the same? There isnt a skew towards the red/blue scanline in its size/falloff etc. I will get to the bottom of this issue though but I think I need to get the spectrometer to debug it properly. Ill have a look for those you mention.
1 Like
The NTSC encoders may have taken it into account, but the RGB output typically never hits that circuit and has very different colors. If you have an SNES, it’s pretty easy to verify this by comparing the RGB/SCART output with svideo or composite output.
2 Likes
Yeah maybe Ill have to invest in some original hardware. This might be a slippery slope… 
1 Like
@Dennis1 is this spectrophotometer any good do you know? https://www.amazon.co.uk/Calibrite-ColorChecker-Display-Calibration-Creatives/dp/B0973JMM4S
RetroArch’s internal HDR settings + RetroArch image viewer.
If you look at the primary points, there are also numerous dots well outside the gamut triangle. I think most of the black dots inside the triangle were the result of flawed test images?
This is what i get with an image purportedly containing all 16,777,216 24bit colors once each:
The remaining dots inside the triangle could still be a flaw with the image i suppose, or they could be a clue as to what the problem is.
If we look back at my previous testing, Megatron seems to display the same issue in a different way, with dots clustering around the phosphor primary points well outside our target gamuts.
While we obviously don’t want to exceed whatever our target gamut is, every Colour System/Phosphor/alternate gamut setting includes colors outside of SDR/Rec 709, including Rec 601, so we definitely don’t want to hard lock Megatron to SDR/Rec 709 colors. (Which also touches on the changes i’ve been making to hdr10.h in my gamut fix testing fork, as Megatron currently seems to lock to Rec 709 colors aside from the error colors thing.)
Keep in mind that different devices could have very different results in this regard too.
For example, for the purposes of DOS games, the current emulator gamma output should be correct to the best of my knowledge.
On the other hand, Windows 10/11 HDR currently uses an incorrect decoding gamma for many SDR games when translating them into an HDR container or AutoHDRing them (piecewise sRGB instead of power). (Not sure if this effects the ReShade version of Megatron in general? It looked like you may have done your own gamma decode, but i am also on Win10, so i haven’t been able to test it xD)
Also any adjustment to encoding gamma should probably occur before the NTSC shader for best results?
So might it be best to make a separate, shader specifically for the purpose of correcting gamma encoding when necessary, and keep Megatron solely focused on the CRT end of the chain?
1 Like
This is not a spectrometer but a colorimeter. You rather need something like this:
This is the older model i1 Pro:
The first one seems to be a bargain for 199 pounds, I paid about 500 euros a few years ago for my i1 Pro 2.
And don’t get confused with the name of the first one: The seller wrote the name wrong, the device shown on the pictures is clearly an i1 Pro 2 spectrometer, which normally has not the word “Display” in the product name and is not a Xrite I1 Display Pro, which is the colorimeter you linked to on Amazon.
2 Likes
I didn’t play with the scanline settings yet, as I have yet to learn how to use them properly. I just dialed in the basics like 8K resolution, 800 TVL, r709 color space, some horizontal deconvergence adjustments for red an blue to get rid of the “rainbow” artifatcs which can show up with some resolution / TVL combinations etc.
2 Likes
@Dennis1 The Neutral preset i ported over from the ReShade version of Megatron for you seems to be of a more neutral character than the Default preset. Have you taken a look at that one yet? (Make sure you switch the Colour System to 709. I left it on 601 since that is what it was set to in the ReShade version.)
1 Like
I also discovered a great way to evaluate the phosphor arrangement of the different mask, resolution and TVL combinations, before taking the display into account. I think this a great way to do some fine tuning of the settings within the shader to get it as good as possible presented on the display later.
For example, this is a screenshot I took within Windows with the reshade version of Megatron with resolution set to 8K and 800 TVL:
With the Windows 11 image viewer I can zoom in really close (up to 5000% magnification) and this is what I get with the Megatron shader:
I find it pretty miraculous how the Windows 11 image viewer presents the Megatron shader in a way, which seems to be a close up of a real CRT tube. I know, that my TV is not able to show the same thing here, but the Megatron shader itself is just fantastic.
You can just save the first image from Daytona USA 2 and zoom in for yourself with a proper image viewer and see how well the shader is capable to imitate all the different phosphors.
I could easily post these close up screenshots on the Reddit CRT forum and nobody would think that this comes from a CRT shader 
I believe, that an 8K display with a RGB subpixel layout would get pretty close to what we are seeing here on the screenshots.
4 Likes
Hi @azurfel just had a play about with the reshade lilium hdr analysis and yes I see the colours outside the rec. 601/709 space so yes hmm something odd is going on.
Here are my photos - best not to look at colours in the image as my camera is skewing them but the graphs are valid.
1 Like
Ok my guess at the moment is that this is due to the scanlines. Im guessing Im pumping up the values beyond the original values or something.
I think its a case of just stripping out parts of the shader at time and seeing what fixes/breaks it. When I find the time again Ill investigate it.
2 Likes
Of course 🤦 Good job I checked with you! Mind you spectrometers are blinking expensive!
1 Like
Yes that does seem to be an absolute bargain from what ive seen. Hmm tempted - maybe I can get the wife to buy me an early Christmas present.
1 Like
That is really nice - it really takes the edge off those pixelated phosphors. We need more pixels thats for sure.
I do wonder whether my fall offs are working as I expect - mind you its a minor thing and maybe related to what @azurfel has found with overly bright pixels and youve found with the off whitepoint. Itd be really nice if they were all related to one bug (in reality itll probably be due to 10 different ones )
1 Like
Daytona is blinking good looking game isn’t it.
1 Like
Note that RetroArch HDR without Megatron active does it too. Take a look at that test image i linked with all the 24-bit colors in RA’s image viewer.
Doesn’t mean the scanlines aren’t a factor in the Megatron variation of the issue, of course. Like you said, could be 10 different bugs xD
Oh, worth noting that the alternate hdr10.h gamuts i’ve been playing around with just move the scattered pixels around along with the phosphor primaries. The pattern in which they cluster around the primary seems to remain the same.
2 Likes
Ah yes - you did say that! 🤦 Well that narrows down the problem considerably! Hmm well Ill take that as good news - quartered the work in one fell swoop.
2 Likes