You can't compare LCD PPI to OLED PPI because most OLED phone displays are pen tile. The pixel density OLED displays have to be much higher to overcome the screen door effect.
Galaxy Note 9 - 516 x 0.75 = 387ppi
iPhone XS - 458 x 0.75 = 343ppi
iPhone XR - 326 x 1 = 326ppi
I'm not sure this kind of math hold :unsure:
The issue is of course colors aren't equal to human eyes, and so subpixels are not a really good unit compared to pixels, but I let you correct me if I'm wrong when trying to go at the sub-pixel level:
This is the Pentile Diamond arrangement Apple is using :
I can't find any source of how exactly pixels are considered, but going with this subdivision for 1 point (@3x then 9 px):
With then 9G per point and either 4R+5B or 5R+4B, it gives:
Pentile Diamond 458 ppi @3x
1 pt == 9 px == 18 subpx (4.5R + 9G + 4.5B)
2*458/2.54 =
361 subpx/cm to continue with linear units
but more logically, a pixel should be related to area units:
2*458^2 = 419528 sppsi (subpixels per square inch)
419528/2,54^2 =
65027 subpx/cm^2
And so
32513 G and
16257 R or B subpixels per cm^2
and we verify √(65027 * 2) = 361
But when looking at the density of data on screen,
with 18 subpx per point : (458/2.54)^2 / 9 = 65027/18 =
3613 pt/cm^2
To compare with:
RGB 326 ppi @2x
1 pt == 4 px == 12 subpx (4R + 4G + 4B)
3*326/2.54 =
385 subpx/cm
3*326^2 = 318828 sppsi
318828/2.54^2 =
49418 subpx/cm^2
And so
16473 R, G or B subpixels per cm^2
and √(49418 * 3) = 385
And with 12 subpx per point : (326/2.54)^2 / 4 = 49418/12 =
4118 pt/cm^2
And for reference:
RGB 264 ppi @2x
1 pt == 4 px == 12 subpx (4R + 4G + 4B)
3*264/2.54 =
312 subpx/cm
3*264^2 = 209088 sppsi
209088/2.54^2 =
32409 subpx/cm^2
And so
10803 R, G or B subpixels per cm^2
And with 12 subpx per point : 32409/12 =
2701 pt/cm^2
RGB 401 ppi @3x-scaled 1242x2208->1080x1920
115x115 pt => 9x100x100 px == 270000 subpx
1 pt => 270000/13225 = 20.4 subpx
3*401/2.54 =
474 subpx/cm
3*401^2 = 482403 sppsi
482403/2.54^2 =
74773 subpx/cm^2
And so
24924 R, G or B subpixels per cm^2
But with 20.4 subpx per point: 74773/20.4 =
3662 pt/cm^2
And conclusion ...

well, anyway most people don't see any real difference in sharpness between these screens
But it comes the Pentile screens Apple is using are 32% sharper than the screens on their non-Plus iPhone (65027 vs 49418 subpx/cm^2), disregarding colors matters (same number of R and B subpixels, but double the number of G subpixels on the OLED screen).
And the iPhone Plus models are 15% sharper than the X models (74773 vs 65027 subpx/cm^2).
Now, the OLED screens and the iPhone Plus screens have about the same density of data when the 326 ppi RGB screens are denser, and this translates for the latter to elements displayed smaller on screen and then with fewer subpixels.
So to me the 458 ppi Pentile diamond screens appear definitely superior to the 326 ppi RGB screens in term of sharpness. And the issue with the sharper 401 ppi RGB screens is the necessary scaling that will display blurry horizontal or vertical 1 pt lines, and messing with scrolling.
Well if you look at the subpixels, Red and Blue have pretty much the same number as the XR. This can't be a coincidence.
Red 324 SPPI
Green 458 SPPI
Blue 324 SPPI
Where is this coming from? I understand that if there is 458 green subpixels per inch horizontally or vertically, with a Pentile diamond arrangement, it's only diagonally that there is 458/√2 = 324 red or blue subpixels per inch.
There is only 2 subpx per pixel, and 1/2 of the subpx are green; I don't think you can do 2.54*√16257 = 324 and 2.54*√32513 = 458 to get the R/B and G subpixels count per inch, basing on previous math.
[edit]The linear units show issues when you want to manipulate subpixels basing on pixels, depending on the fact you're in RGB or with a Pentile screen, but you don't face this kind of problem with points of course:
For example 2.54*√2701 = 132 pt/inch and 2.54*√4118 = 163 pt/inch (or more simply: 264/2 and 326/2 because each point is composed of exactly 2*2 pixels)
And the point size on the Pentile screen and the Plus screen is equivalent, and fall between the two others: 2.54*√3613 = 153 pt/inch and 2.54*√3662 = 154 pt/inch (or 458/3 and 1.15*401/3)
[/edit]
Shouldn't you rather compare the green density with 458*0.5 = 229 red or blue subpixels per inch horizontally or vertically?
Because otherwise, if you want to look at the diagonal density of subpixels, there is then 648 green subpixels per inch. But I don't think you can consider that pixels are turned 45° like follows:
And I don’t think either Apple is doing complex stuffs like this:
where every pixels have 4 subpixels but one subpixel is shared with every contiguous pixel. We’d get:
1 pt == 9px == 24 subpx (6R + 12G + 6B)
458 px per inch => 4 + 3*457 = 1375 subpx per inch
hence 1375/2.54 =
541 subpx/cm
458^2 px per sq in => u(458) = 420444 sppsi
[ with u(1) = 4 and u(n) = u(n-1) + 4n = 2*n*(n+1) ]
hence 420444/2.54^2 =
65169 subpx/cm^2
And so
32584 G and
16292 R or B subpixels per cm^2
but with 24 subpx per point, we get: 65169/24 =
2715 pt/cm^2
and unless I'm making a mistake, this last value should be close to the one found with the iPhone Plus screen, as seen before, given the size of the elements displayed on the OLED Pentile screen iPhone models (but we can note this pixels arrangement could be a way to emulate the iPad with 264 ppi RGB displays on a 458 ppi Pentile diamond screen. Another solution being to use @3x a 401 ppi RGB panel for example, then with 27 subpx per point to get 74773/27 = 2769 pt/cm^2 , given 132*3 ~ 401).
edit: there is one mistake or imprecision in the last calculations above:
When the screen is seen as a 458 ppi screen with 2 subpixels per pixel, we found 361 subpx/cm.
If it is now seen as a 458 ppi screen with 4
shared subpixels per pixel, we probably should rather have 4 + 3*457
- 1 = 1374 subpx per inch (but of course, at this size, the width of a subpixel is well inside the margin of error).
This still gives 1374/2,54 =
541 subpx/cm anyway
(and we can see here the problem when using linear units: one has to count here the number of subpixels contained in a 1 pixel thick line. And that’s why, despite being exactly identical screens, we find 361 ≠ 541 subpx/cm that are both valid results)
Similarly, there is u(458)
- 2*458 = 419528 sppssi and now we find the same value as before, logically (was bugging me)
And so giving the same
65027 subpx/cm^2
still with
32513 G and
16257 R or B subpixels per cm^2
And finally, with 24 subpx per point, we now get: 65027/24 =
2709 pt/cm^2
and the rest of the post hold wrt to the iPad 264 ppi RGB screens equivalence.