Guest-Advanced 4k HDR1000 experiments

RetroArch Additions Shaders
#17

Brightness Tests

Here are some measurements taken with a light meter, full white, 100% window, no scanline mode. Average of 3 readings, light meter placed against the screen. Round to the nearest whole number. Measurements in nits.

Mask			Nits

MGX:			412
RRGGBBX: 		350 
MGX+slot:		306
dotmask: 		300
RGBX:			300
RRGGBBX+slot:	263
RGBX+slot:		223

These readings agree nicely with my eyeballings of the masks in my previous post. I think the minimum for CRT-like brightness on a sample and hold display is around 300 before scanlines are applied. RRGGBBX+slot is just barely out of reach- the brightest I can get it is around 275 by fiddling with HDR settings. Maybe it’s enough in a dark room? This also means that RGBX+slot is also ruled out.

Scanlines actually consume a bit more brightness than I thought - here are the readings with my scanline profile applied:

Mask						Nits

MGX+scanlines:				314
RRGGBBX+scanlines:			264
MGX+slot+scanlines:			236
dotmask+scanlines: 			229
RGBX+scanlines:				225
RRGGBBX+slot+scanlines:		201
RGBX+slot+scanlines:		170

For reference, a high-end consumer CRT could achieve a maximum brightness of anywhere from 200-300 nits (full-screen white image). However, this was only briefly sustained and would drop slightly.

It’s important to also acknowledge that perceived brightness on a CRT was likely higher than a sample and hold display of the same nits, due to variable persistence of the electron beam. John Carmack did a talk at the 1st Oculus connect where he talked about the effect of persistence on perceived brightness.

It would be interesting to see if beam simulator on a high Hz display has an effect on perceived brightness, but my monitor only goes up to 160Hz.

3 Likes
#18

It’s hard to pick a favourite but I feel the least connection to Trinitron Masks, but I do love the scanline dynamics they produce at higher TVLs.

I’ve also seen the brightest masks with the most vibrant colours with Trinitron Masks in Sony Megatron Colour Monitor.

When I used CRT-Guest-Advanced via Mega Bezel, my Shadow (Dot) Mask presets ended up the brightest with slot mask all else equal being the darkest.

I then attempted to normalize brightness among all 3 mask types in my Neo-GX preset pack.

Dot mask is starting to become my favourite for various practical as well as subjective reasons.

Firstly, the dynamics are almost as good as Trinitron. On our digital displays the roughness that the dot triads add to the scanline help make them look more "analogue’ to me.

It’s a bit like deconvergence, when it’s completely disabled, the phosphors and scanlines look a bit robotic.

I also tried using Fringing and Artifacts and found that they work well in many scenarios and also contribute to making the signal feel more analogue. I got used to fringing after a few minutues.

They don’t work when I’m trying to fake additional colours and transparency for Turbo Duo games though as they add err artifacts to the areas where there should be just new colours.

That’s the main reason for my 3D Comb Filter presets.

Slot Mask is probably challenging because of the extra darkness potential but it looks great once you get it right. I think I accomplished this with my Old Skool Arcade Epic Presets. I almost always use Slot Mask with Arcade games for authenticity but dot mask looks awesome here as well.

1 Like
#19

I know what you mean - low TVL slotmask is probably my ideal for most retro systems, but the brightness is just barely not there - HDR1400 should definitely get us the rest of the way there. It’s also possible that the perceived brightness of True Black HDR1000 will be higher than regular HDR1000.

It’s sort of “maybe adequate, takes a lot of work” vs “clearly works” for the ones I’ve recommended.

I haven’t made any attempt to do this since it would mean introducing clipping - I’m probably one or two clicks away from clipping something with my current configurations.

I’m also not using any brightness mitigations - even a single click on some things can yield a pretty significant gain in perceived/actual brightness- after you loosen that restriction, a lot more possibilities open up.

I like RRGGBBX and dotmask configuration about the same, I flip between the two depending on the mood. I think you’re right that the dotmask feels a bit more analogue, but RRGGBBX is quite a bit lower TVL, which has its own benefits - sometimes less is definitely more when it comes to pixel art.

I see you’ve been using a similar dotmask in some of your presets, looks good. As I recall reading, some of the earliest color TVs used a low TVL dotmask (Zenith Chromaclear)- this should probably be paired with a very smeary and blurry composite video signal for authenticity :stuck_out_tongue:

Yep, sometimes when you disable these completely it feels a little too perfect, and some things can look a bit better with a bit of fringing/artifacting. Artifacting 0.00 / Fringing 1.00 is probably as good as it gets for a CRT. Sometimes I like to crank things up a bit higher because it just feels right.

Yep, quite a bit - adding the slotmask is a ~25% loss of brightness according to my calculations and confirmed by tests.

I can get dotmask a lot brighter if I use a “standard” configuration- CRT Mask 4.00, Lottes MaskDark 0.00, MaskLight 1.00

This measures 296 nits using the same test settings as my above tests.

Since the RGB triad is 4 pixels wide at the widest point, that’s 540 TVL.

I’m using an alternative dotmask- Mask 10.00, (Transform to) Shadow Mask 1.00

This adds a black pixel, which prevents the red and blue phosphors from blurring together and helps with phosphor shape. Obviously, it’s a bit darker due to the addition of the black pixel.

The TVL is 432, so it’s a bit closer to consumer grade and has a bit more texture.

Taking another look at the Lottes configuration, it’s not that bad if you don’t mind sacrificing the phosphor shape a bit and don’t mind your reds and blues blurring into magenta. The extra brightness is definitely nice. It does need a bit of extra work to fine-tune, though, compared to the others- an HDR adjustment might be enough (needs a bit more contrast)

1 Like
#20

You can try RGB Mask 6 for dotmask then manipulate the size. That’s what I did for my Neo GX presets and when I was trying to match my CRT-Guest Advanced presets to My Megatron presets.

1 Like
#21

Looks to be the same as the Lottes layout above but without the weird color/brightness/etc shift.

1 Like
#22

Update:

-Replaced rgb presets with new presets using the ntsc version of the shader, tweaked for perfect sharpness/color/etc. The additional bit of processing looks better than the regular version of the shader- it’s actually capable of being a bit sharper with lower TVL masks.

-Spent quite a bit of time tweaking the horizontal filter- this should be the safe maximum for sharpness. Let me know what you think- best judged with a high TVL mask like MGX.

-Gamma corrected (standard 2.5/2.2) - helped color separation in highlights and improves the colors

-Adjusted beam profile accordingly. Try increasing Scanline Shape Bright Pixels and you’ll start clipping almost immediately. Try reducing it and the scanlines disappear over the brightest colors. Adjusted the dark beam as well to compensate for gamma adjustment.

-Added slight vertical deconvergence- I think it improves shade separation in highlights, lets you eek out a bit more brightness. Basically just gives you a bit more scanline variability. Plus, it looks good :slight_smile:

-To add slotmask, simply adjust “Slot Mask Strength Bright Pixels” and “Slot Mask Strength Dark Pixels” to 1.00. Only recommended if you have HDR1400…

-No more separate NTSC presets - instead, this handy cheat sheet can be used to adjust the parameters to whatever look you prefer :smiley:

Guest-Advanced-NTSC Shader Preset Cheat Sheet

2 Likes
Please show off what crt shaders can do!
#23

Ok, quick update-

After some additional tests, gamma adjustment really should be done with the HDR settings when using HDR and left neutral in the shader (e.g., 2.4/2.4)

My original recommendations for HDR settings are mostly still good- Peak Luminance 1000, Paper White 750. Contrast should come down to ~5.00.

2 Likes
#24

I gained enough brightness with the recent changes that RRGGBBX+slot is back on the menu!

Measured 212 nits- comfortably above my target of 200 nits :smiley:

No clipping involved. Just adjust Paper White Luminance to 1000.

2 Likes
#25

Can you share your measurement app again please?

Are you referring to Peak and Paperwhite Luminance settings as well as Contrast?

2 Likes
#26

It’s a light meter I picked up for $30 on Amazon from an alphabet soup Chinese brand. URCERI

Does the trick!

Yep, although ideally Contrast should be left alone (5.00) to avoid any clipping. A couple notches might be ok, up to 5.50 or so, maybe 5.6 max (with the darkest masks). Those have been my findings so far.

2 Likes
#27

Ok, so now for the downside:

Cranking up the Paperwhite Luminance and Contrast does soft clip the colors- the shader’s scanlines and mask keep the shades separate but they’re all the same RGB values.

The bigger problem, though, is that when we increase Contrast or Paper White we’re raising the black level, and we’ve already destroyed the black level by disabling local dimming and blasting 1000 nits :joy:

This of course ruins the contrast ratio and has a very negative effect on perceived brightness.

So, slotmask is still “use at your risk” - we really probably want True Black HDR1000 (or whatever it’s called), so give it another 5 years.

1 Like
#29

A quick local dimming experiment:

Enabled local dimming. Used 240p Test Suite’s white screen to set mask strengths, in conjunction with a light meter. Targeted 220 nits, equalizing the peak brightness with my regular preset. This resulted in:

 maskstr = "0.850000"
 mcut = "0.800000"

I also had to adjust gamma in/out to 2.4/2.4 from the usual 2.4/2.2.

Local dimming on, reduced mask strength:

Local dimming off (full brightness), full mask strength:

You can see the horrible backlight bleed when local dimming is off. Honestly, I’m having a hard time deciding which I prefer more. The true black looks so good, and is such an essential aspect of CRT performance, but it comes at the cost of lost mask detail. In practice, the lost detail/accuracy isn’t very noticeable when playing a game, so I might prefer the local dimming. IDK, it’s a tough call.

2 Likes
#30

Yes, use Local Dimming. You paid for it, didn’t you? The problem is, you’re using an IPS display which has horrible native black levels and contrast ratio as well as blooming performance so it absolutely has to rely on local dimming to get anywhere near decent contrast or black levels, however local dimming algorithms vary considerably so there’s the potential to introduce a different set of artifacts due to that.

So there are times when my TCL QM8751G gently reminds me that’s it’s only a miniLED and times it reminds me that it’s not an OLED but those times are actually in the minority. Most of the time I’m very satisfied with what I have and I know what other frustrations I would have to deal with and things I’ve gotten used to that I might miss out on if I go back to RRBBGGX only WOLED.

So we’re getting closer to an R-G-B tandem with no white subpixel, hopefully with MLA sometime in the future and that is when the OLED proposition becomes much more palatable for those who would have the unlimited loads of cash to burn of course, because you see how they’re pushing the brightest OLEDs with the best technology as ultra premium devices.

So that’s out of the question for me. So I’ll happily stick with plebian miniLED tech. VA is where it’s at though with the awesome native black levels which is more than enough. I can’t turn local dimming off on my TV because the highest brightness levels are tied to Local Dimming being on High, but if you ever experienced a TCL QM851G you probably wouldn’t want local dimming turned off either.

So even if you’re not going to purchase one, do take some time and study the TCL QM851G, QM8K and QM9K. The QM751G is just a baby QM851G with 720 Zones on a 55" Screen with a peak brightness of I think about 2,400nits!

Next year will be the year of RGB miniLED with much higher gamuts and brightness. Lets see how that goes.

PS, once you start using all of the features of your IPS miniLED, you’ll get used to them eventually and adjust your shader parameters to suit them best. I made the most out of an old crappy LG IPS by trying to run it at the peak of it’s brightness and saturation abilities, why? If black levels were poor and I was going to get the most out of that TV, I had to focus on brightness to achieve the highest contrast.

HDR video? No problem, I reverse engineered my own solution to watch HDR videos which I had gotten accustomed to on my non-officially HDR display!

By the way, after coming across some further information concerning my Cable Matters DisplayPort 1.4 to HDMI 2.1 Adapter I have managed to eek out RGB 4:4:4 8-bit at 100Hz with HDR and VRR enabled. Banding in gradients should be kept to a minimum via the use of Massoft’s Color Control, which allows me to enable nVIDIA’s excellent temporal dithering algorithm on Windows 11.

So this is the theoretical max refresh rate you can get out of an Nvidia 10 series GPU while maintaining RGB 4:4:4.

Anything higher requires HDMI 2.1 which my graphics card lacks or Display Stream Compression which it also lacks.

Not to be satisfied, I’ve recalibrated my HDR experience with BFI enabled and I have gotten some pretty interesting results. So I’ve now begun my high framerate journey and I’ve started using BFI again!

Meanwhile… https://www.reddit.com/r/buildapcmonitors/s/NLnjeGRzMS

1 Like
#31

It was fun for about a day, then the high wore off and I started to notice all of the problems.

The problem is that entire colors are dimmed, and this happens to already dim colors.

I think the problem is with the mask and scanlines adding too many black pixels to dark colors, confusing the algorithm and making it think “this should be black.”

The algorithm works very well for very high contrast content - like old black background games with a few bright colors. It’s the midtones that get screwed up.

So, yeah… True Black HDR1000 can’t get here fast enough.

1 Like
#32

I understand, not all miniLED backlight algorithms are the same though and more dimming zones also help with the brute forcing. The problem with the IPS miniLED is its over reliance on the local dimming technology to achieve an acceptable level of contrast. Due to VA’s superior native contrast ratio, a VA miniLED might not need to be so aggressive.

Also, on my display, I can tailor what gets prioritiezed by choosing a different tonemapping setting. For example Detail Preferred, Balanced, Brightness Preferred and High-Dynamic.

Maybe it’s time to trade up for a miniLED with more modern features and technology?

To summarize, local dimming performance can vary considerably between different brands and models across the miniLED landscape and this variance can range from excellent - near OLED performance to practically unusable - just a checkbox to tick.

1 Like
#33

Maybe, but this display is no slouch- It’s basically just a rebranded version of this:

I’m not super confident that the dimming issue I’m describing can ever be completely resolved, but a less aggressive dimming and/or more dimming zones might help reduce it.

True Black HDR1000 will save us all, hopefully. I’m a bit worried that it will remain restricted to small screen devices.

2 Likes
#34

By the way, have you been able to access similar firmware updates to the Cooler Master?

1 Like
#35

I’ve been meaning to do that- might give that a shot a later today.

1 Like
#36

Seems the download was removed?

1 Like
#37

Thanks, should be working now :smiley:

1 Like