4 ms·

Diamond Subpixel Arrangement Simulator

92 points by firloop 3y ago

m3kw9 3y ago
[flagged]
- wahnfrieden 3y ago
  No
- notso411 3y ago
  What?
- jxy 3y ago
  That's what would happen if you use Planck pixel renderer.
- chucky 3y ago
  I'm not sure if you are joking or not, but in case you aren't, your iPhone won't let you directly address subpixels, so no. There's a whole subcategory of beautiful bugs caused by recursive algorithms going way farther than their authors intended - it probably doesn't apply here.
tomsmeding 3y ago
That 4K render of an iPhone screen is so uselessly accurate it's great.
- matthewfcarlson 3y ago
  I’ve seen plenty of photoshopped on iPhones. This one really did look more real which I didn’t expect
DiabloD3 3y ago
Neat, but I kinda wish manufacturers would quit making cursed displays like this. Either follow the existing convention, or try your best to hide the subpixels entirely.
- fidotron 3y ago
  This layout has the distinct advantage you can rotate the screen with everything remaining the same (font hinting etc.) Varying subpixel layouts are one of the major reasons I struggle to take the obsessions of "pixel perfect" designers seriously [1]. It is the kind of notion you can only advocate if you're ignorant of the reality of what is going on. Display aging and resulting color inaccuracies make this even more fun. [1] Not helped by having worked on products where the designers quoted 25 years (not a typo) as the time required to support all the supposedly necessary profiles, and the resulting build size would have been too big and so we lost a multi million dollar opportunity. Of course anyone opposing this was simply an enemy of quality. Edit to add: for those unfamiliar https://en.wikipedia.org/wiki/ClearType https://en.wikipedia.org/wiki/ClearType was a very sneaky way to take advantage of the fact that the subpixel layout dramatically alters the outlines the viewer sees.
  - infotainment 3y ago
    Do people still worry about “pixel perfection”? It made sense in the era of low resolution displays, but with everything super high PPI these days, it seems like it typically doesn’t matter.
    - leduyquang753 3y ago
      Many computer screens are still at 1080p, including my 15-inch laptop which resolves to only 142 PPI.
  - DiabloD3 3y ago
    It's actually because of ClearType that I hate screwing with layout. This doesn't really effect the iPhone, but a lot of apps completely ignore your intended subpixel layout. Windows or X/Wayland can be set correctly, and many apps do handle that correctly, but many don't. Many enable subpixel hinting when it is clearly disabled by the user. Want greyscale because your monitor fringes hard? Too bad. Want aliased because it is a low resolution display or you hate fuzzy fonts? Too bad. Also, many apps do detect correctly, but only detect your primary monitor, not the monitor the window is on. Doesn't effect you if you're single monitor. And to further shit on broken apps: there are apps that are already using DirectWrite on Windows, and are force overriding the correct (default) settings. Why?!?! You were so close! And to shit on the Internet as a whole: people screenshotting text for their websites that have Vista-era Cleartype tuning clearly in it. Isn't so bad on RGB monitors, but almost unreadable on BGR or any V orientation. And one last insult: DirectWrite and Freetype only support RGB, BGR, VRGB, and VBGR. There are many other weird ass configurations, such as LG C2s having RGBW (which causes extreme haloing on text), and another OLED panel (forgetting which monitor) has RBG (why?!). So screw it. I won't use non-RGB monitors until we live in a world we're doing 300% DPI equivalent (ie, 5760x3240 in a 24"; best I can get in a commodity is 4k in 24", or 200%), and I can't see the subpixels anymore no matter how much I squint.
Retr0id 3y ago
The colour tinting of the mac icon and folder examples implies that their code isn't accounting for gamma curves properly, I might see if I can contribute a fix.
- RickyRomero 3y ago
  The screenshots shown in my post were from an older version. The current version should work correctly. The purple tint was coming from naïvely multiplying the red and blue channels by 2 to try and compensate for the fact that you have half as many subpixels of R/B to work with as compared to the green channel. My new approach cuts the maximum green brightness (after converting to linear gamma) instead. https://rickyromero.com/tmp/diamond-happy-mac.png https://rickyromero.com/tmp/diamond-happy-mac.png I stood pretty far back from my laptop's screen and it looks more accurate to me now :)
  - Retr0id 3y ago
    Ah, good to know, I can see the gamma compensation in the green division. fwiw you should probably also convert to linear before doing the bilerp, and back afterwards, but that matters comparatively less.
  - Affric 3y ago
    This is a wonderful post. I cannot stop gazing at the mundane beauty.
zamadatix 3y ago
This seems to treat one white dot as a green subpixel which bleeds 25% into the diagonally 4 adjacent red/blue subpixel. I've always seen pentile subpixels represented as groups of 2 element subpixels https://geometrian.com/programming/reference/subpixelzoo/square_pentilediagonalGRGB.png https://geometrian.com/programming/reference/subpixelzoo/squ... or diamond shaped gorups of 4 subpixels, 2 of which are green https://global.samsungdisplay.com/wp-content/uploads/2021/11/Untitled-design-2021-11-19T163245.312.png https://global.samsungdisplay.com/wp-content/uploads/2021/11.... In either case, the green subpixels tend to be smaller in response to their increased prevalence.
mikaelaast 3y ago
I love this. I tried doing sub-pixel simulation for a tool I created (screenstab.com if anyone’s interested – yeah I know, shameless plug, etc.). I ended up abandoning the sub-pixel aspect in my shader because of the distracting patterns caused by the Moire effect.

92 points by firloop 3y ago

m3kw9 3y ago

[flagged]

wahnfrieden 3y ago
No
notso411 3y ago
What?
jxy 3y ago
That's what would happen if you use Planck pixel renderer.
chucky 3y ago
I'm not sure if you are joking or not, but in case you aren't, your iPhone won't let you directly address subpixels, so no. There's a whole subcategory of beautiful bugs caused by recursive algorithms going way farther than their authors intended - it probably doesn't apply here.

tomsmeding 3y ago

That 4K render of an iPhone screen is so uselessly accurate it's great.

matthewfcarlson 3y ago
I’ve seen plenty of photoshopped on iPhones. This one really did look more real which I didn’t expect

DiabloD3 3y ago

Neat, but I kinda wish manufacturers would quit making cursed displays like this. Either follow the existing convention, or try your best to hide the subpixels entirely.

fidotron 3y ago
This layout has the distinct advantage you can rotate the screen with everything remaining the same (font hinting etc.) Varying subpixel layouts are one of the major reasons I struggle to take the obsessions of "pixel perfect" designers seriously [1]. It is the kind of notion you can only advocate if you're ignorant of the reality of what is going on. Display aging and resulting color inaccuracies make this even more fun. [1] Not helped by having worked on products where the designers quoted 25 years (not a typo) as the time required to support all the supposedly necessary profiles, and the resulting build size would have been too big and so we lost a multi million dollar opportunity. Of course anyone opposing this was simply an enemy of quality. Edit to add: for those unfamiliar https://en.wikipedia.org/wiki/ClearType https://en.wikipedia.org/wiki/ClearType was a very sneaky way to take advantage of the fact that the subpixel layout dramatically alters the outlines the viewer sees.
- infotainment 3y ago
  Do people still worry about “pixel perfection”? It made sense in the era of low resolution displays, but with everything super high PPI these days, it seems like it typically doesn’t matter.
  - leduyquang753 3y ago
    Many computer screens are still at 1080p, including my 15-inch laptop which resolves to only 142 PPI.
- DiabloD3 3y ago
  It's actually because of ClearType that I hate screwing with layout. This doesn't really effect the iPhone, but a lot of apps completely ignore your intended subpixel layout. Windows or X/Wayland can be set correctly, and many apps do handle that correctly, but many don't. Many enable subpixel hinting when it is clearly disabled by the user. Want greyscale because your monitor fringes hard? Too bad. Want aliased because it is a low resolution display or you hate fuzzy fonts? Too bad. Also, many apps do detect correctly, but only detect your primary monitor, not the monitor the window is on. Doesn't effect you if you're single monitor. And to further shit on broken apps: there are apps that are already using DirectWrite on Windows, and are force overriding the correct (default) settings. Why?!?! You were so close! And to shit on the Internet as a whole: people screenshotting text for their websites that have Vista-era Cleartype tuning clearly in it. Isn't so bad on RGB monitors, but almost unreadable on BGR or any V orientation. And one last insult: DirectWrite and Freetype only support RGB, BGR, VRGB, and VBGR. There are many other weird ass configurations, such as LG C2s having RGBW (which causes extreme haloing on text), and another OLED panel (forgetting which monitor) has RBG (why?!). So screw it. I won't use non-RGB monitors until we live in a world we're doing 300% DPI equivalent (ie, 5760x3240 in a 24"; best I can get in a commodity is 4k in 24", or 200%), and I can't see the subpixels anymore no matter how much I squint.

Retr0id 3y ago

The colour tinting of the mac icon and folder examples implies that their code isn't accounting for gamma curves properly, I might see if I can contribute a fix.

RickyRomero 3y ago
The screenshots shown in my post were from an older version. The current version should work correctly. The purple tint was coming from naïvely multiplying the red and blue channels by 2 to try and compensate for the fact that you have half as many subpixels of R/B to work with as compared to the green channel. My new approach cuts the maximum green brightness (after converting to linear gamma) instead. https://rickyromero.com/tmp/diamond-happy-mac.png https://rickyromero.com/tmp/diamond-happy-mac.png I stood pretty far back from my laptop's screen and it looks more accurate to me now :)
- Retr0id 3y ago
  Ah, good to know, I can see the gamma compensation in the green division. fwiw you should probably also convert to linear before doing the bilerp, and back afterwards, but that matters comparatively less.
- Affric 3y ago
  This is a wonderful post. I cannot stop gazing at the mundane beauty.

zamadatix 3y ago

This seems to treat one white dot as a green subpixel which bleeds 25% into the diagonally 4 adjacent red/blue subpixel. I've always seen pentile subpixels represented as groups of 2 element subpixels https://geometrian.com/programming/reference/subpixelzoo/square_pentilediagonalGRGB.png https://geometrian.com/programming/reference/subpixelzoo/squ... or diamond shaped gorups of 4 subpixels, 2 of which are green https://global.samsungdisplay.com/wp-content/uploads/2021/11/Untitled-design-2021-11-19T163245.312.png https://global.samsungdisplay.com/wp-content/uploads/2021/11.... In either case, the green subpixels tend to be smaller in response to their increased prevalence.

mikaelaast 3y ago

I love this. I tried doing sub-pixel simulation for a tool I created (screenstab.com if anyone’s interested – yeah I know, shameless plug, etc.). I ended up abandoning the sub-pixel aspect in my shader because of the distracting patterns caused by the Moire effect.