POST A COMMENT

152 Comments

Back to Article

  • sciwizam - Friday, October 21, 2011 - link

    /waits for the Techmeme section of all the Apple news sites featuring the 2nd chart. Reply
  • zorxd - Friday, October 21, 2011 - link

    why do we even care about these charts?

    What we care is the total number of pixels. Not the density.

    You wouldn't want a 400 PPI 320x480 display on your cell phone.
    Just like you prefer that 23" 1920x1080 desktop display over a 300 PPI 1024x768 one.
    Reply
  • zorxd - Friday, October 21, 2011 - link

    Think about it. If you want a very high number of pixels per inch, just move your phone 1 meter away from your eyes. This way you won't see the pixels. You can hold a 4.65" display farther than a 3.5" display by the way.

    This just prove are irrelevant the PPI calculation is.
    Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    All pixel-based displays strive for their viewers to be unable to distinguish individual pixels.

    Different PPIs are required to achieve the effect at different distances.

    If I have to get 4' away from my 1080p HDTV to discern individual pixels, then that's a good thing. Most times, I will use the TV much further away, but I might walk closer. I don't want the visual quality to suddenly deteriorate if I happen to get up look at it as a I walk past the set.

    The same reasoning is used in mobile phones. I typically use my phone at a distance of 12-18", but the device will undoubtedly get closer to my face at some point during usage.

    So it comes down to usage distance and PPI. Screen size and pixel count, alone, have little influence.

    In marketing, screen size can be used because customers expect a satisfactory PPI based on the usage distance of the product.
    Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    I don't want the visual quality to suddenly deteriorate if I happen to look at the set as a I walk past it.******************************* Reply
  • deV14nt - Saturday, October 22, 2011 - link

    That's actually all we are measuring here.

    I just came to the realization that if you have a 3.5" screen, you probably hold your phone closer than someone who has a 4.65" screen. And as you get closer, you need more PPI to compensate.

    So if you wanted to do more than impress someone who holds the phone right up to their eye, and actually make small text smooth at the actual distance they hold the phone, you would have to know what that distance is. So far, I haven't seen anyone mention anywhere that vital missing piece of information.

    Do iPhone users hold their phones closer than those with bigger screens to see the much smaller text or other elements on the display? How much does that affect the perceived quality at the 2 different distances?
    Reply
  • a5cent - Saturday, October 22, 2011 - link

    Well, kind of, but not really. The issue you raise is related to two things (a) the distance at which a persons eyesight is comfortably focused and (b) the size of UI elements on the screen. The former is more of an issue for older people or those with impaired eyesight.

    Given the same text size, people will tend to hold the device at the same distance, irrespective of display size.
    Reply
  • deV14nt - Monday, October 24, 2011 - link

    But you won't often see the same text size. The most consistent UI element would be the way the web browser is used. For example, displaying almost a full-page view in portrait and a full-width view in landscape. In that common scenario, the text on the 3.5" display will be much smaller than a 4.65" display, so the user will hold the 3.5" display closer, negating some of the benefit of higher pixel density.

    Many apps will share similar layouts and will simply adjust the text size to fit the display as well. So you should see the same thing there in most cases.

    As I said, without knowing the average distance people hold these devices in these very common, reproducible scenarios, the Effective value of their respective pixel densities is unknown. Which is really the only attribute that matters.
    Reply
  • a5cent - Tuesday, October 25, 2011 - link

    You are right. Font size is not consistent across applications. However, we were not discussing font size variance across apps but across display sizes.

    You asked if iPhone users hold their phones closer. The answer is that they do not (given a similar text size).

    You will find many android devices with large displays, where the texting and e-mail applications render text at about the same size as on an iPhone.

    A web page that is comfortably read in full-width mode on a large android phone, will often force an iPhone user to zoom in on the are of interest and scroll/read through an article that way (a mobile browser will apply a level of magnification that attempts to render the text using a predefined font size that is comfortably read on the given display).

    You don't need to know the viewing distance at which you are comfortable reading the average font size used in print media (and smartphones), though it is very good to know the minimal PPI required so you don't notice pixelation at that viewing distance. Given that information, it is easier to choose a device that offers an enjoyable viewing experience at any display size (where bigger isn't always better).
    Reply
  • zorxd - Friday, October 21, 2011 - link

    the thing is, you can hold a 4.65" display farther away from your eyes than a 3.5" display

    Would you change your 1920x1080 50" HDTV for a 20" 1280x720? I don't think so, even if it would be an increase in pixel density.
    Reply
  • OfficerX - Friday, October 21, 2011 - link

    No, but I don't carry my TV in my pocket. Reply
  • rpmurray - Friday, October 21, 2011 - link

    The furthest I can hold a 3.5" display from my eyes is 27". When I try that with a 4.65" display it's still 27". Do you have arms that grow depending on the size of the display? Reply
  • deV14nt - Monday, October 24, 2011 - link

    The furthest isn't the number that matters. The average is the one with significance. Reply
  • zorxd - Monday, October 24, 2011 - link

    Well, if you really hold your phone as far away as possible from your eyes, then I understand why PPI matters to you.

    But for normal people, PPI doesn't matter. Size and resolution do.
    Reply
  • augustofretes - Thursday, November 10, 2011 - link

    You're wrong. If you think people don't see the difference between the iPhone 4 screen and the rest is because you don't want to. DPI is nothing more but a measure of size vs pixels, and it has always been important, since it directly relates to how we perceive a screen... Reply
  • cbutters - Thursday, December 08, 2011 - link

    You just said "PPI doesn't matter. Size and Resolution do...."

    What you are failing to understand is that PPI IS size and resolution. It is just a measure determined by resolution and size.
    Reply
  • deV14nt - Friday, October 21, 2011 - link

    Thanks @zorxd for mentioning that the total number of pixels or subpixels of the whole display, if you demand to discuss only one measure alone, is a better approximation of the display's characteristics. The diagonal alone is 33% bigger on the Galaxy Nexus. These displays are not even in the same class. And it seems really silly that so many respectable tech sites are comparing them as if they are.

    It's like comparing a 40" to a 53" TV, yet all these tech sites seem obsessed with the pixel density, ignoring the fact that it means little without simultaneously looking at the screen size. Apparently we are as obsessed with technical minutiae as we are with holding our phones way too freaking close to our heads. At least some of us, those who can't see the forest for the trees.

    All of this ignores the fact that the contrast ratio for the Galaxy Nexus is 100,000 to 1 compared to the iPhone 4S at 800:1. Again, these screens are not even in the same class. Colors are also more vivid (which I'm sure can be measured), and to balance things out you could have mentioned maximum brightness is sacrificed.

    Instead we get an article with some charts that aren't nearly as relevant as they are made out to be. Why is this not more comprehensive?
    Reply
  • retnuh - Friday, October 21, 2011 - link

    You missed the whole printer DPI battle during the 90s didn't you?

    High DPI or in this case PPI is a good thing, but you're measuring it wrong. You want high PPI in X inches, whether X is 3.5", 4.65", 23", then you figure the actual resolution that would be required to make it happen.

    So using your example a 23" screen at 1920x1080 is roughly 96 PPI. It'd take a 23" display at 5760x3240 to get a PPI of ~287. With resolution independence needed for that sort of high PPI everything would appear razor sharp.

    Your 300 PPI 1024x768 example would leave you with a 4.25" display which as you said isn't a very useful desktop monitor, but in a phone it'd be great.

    This is ignoring all the subpixel stuff from above and assuming a 1:1.
    Reply
  • zorxd - Friday, October 21, 2011 - link

    I agree with you. You always want the highest resolution possible on a given screen size. I prefer a 1280x720 4.65" display over a 800x480 4.65" display.

    But I don't see the point in comparing the density between screens of different sizes.

    A 1280x720 resolution is better than a 960x640 resolution. Whether you prefer a 3.5" display or a 4.65" display is a matter of taste, I guess.
    Reply
  • CPD - Friday, October 21, 2011 - link

    "Burt I don't see the point in comparing the density between screens of different sizes."

    When we're talking about devices that are within a couple of inches of one another, it's not obscene to discuss the density. It makes plenty of sense. Both will be used within an arms length, which means that if one has a lesser pixel density than the other, the pixels will be more visible on the display. Of course, it may not bother some.
    Reply
  • deV14nt - Friday, October 21, 2011 - link

    You're underestimating because of the word "inches." This device is 33% bigger in diagonal than the iPhone 4S. This is a bigger gap than the iPhone 4S being 25% bigger in diagonal than the HTC TouchPro, an old 2.8" Windows Mobile device. That device needed a stylus to poke the right button, and even then it was easy to miss all the buttons crammed into such a tight space. This is just as much a new generation of phones as the iPhone was compared to those old WinMo devices. Yet somehow, Apple got stuck there for 4 years. Glad to know we're still moving forward. Reply
  • zorxd - Monday, October 24, 2011 - link

    Do you complain that your 46" TV has a lower pixel density than your 42" TV? I hope not. Reply
  • a5cent - Saturday, October 22, 2011 - link

    It's not the screen size that is important, but the size of UI elements (especially text). Any one person will prefer a certain font size on their display irrespective of screen size. Given the same font size, an individual will hold devices with different screen sizes at the same distance where the device with the higher PPI will offer the higher quality rendering (ignoring Pen Tile displays for the moment).

    Also, you don't always want the highest resolution possible at a given screen size. At some point humans simply can no longer tell the difference. After that point choosing a device with a higher resolution for a given screen size (PPI) is simply sacrificing performance (pushing many more pixels around) and battery life for no perceivable benefit.
    Reply
  • GuinnessKMF - Friday, October 21, 2011 - link

    Umm... I care about PPI, it depends on what I'm using the device for. I actually prefer my laptop 17" 1920x1200 display over my desktop 24" 1920x1200 display, higher density at the same distance results in better looking images. For photo viewing density is key, for an immersive experience (gaming and movies) then screen size is key.

    There are absolutely uses for this kind of information and I am glad that they provide it, what exactly is the point of saying "why do we care", you may not care, but other people do, just don't read it, what do you gain by it not existing?
    Reply
  • zorxd - Friday, October 21, 2011 - link

    If someone offered you to exchange your 24" monitor for a 15" monitor of 1680x1050 (higher pixel density) would you accept?

    Of course not. Display size matters. And resolution too. Density doesn't, except as a consequence of the first two.
    Reply
  • Skywalker_x86 - Friday, October 21, 2011 - link

    From your argument: you would trade a 20" monitor with 1024x768 resolution for a 36" monitor with the same resolution, which would look much worse by comparison.

    The differences in screen sizes on these phones is minimal, so the density does matter. Pentile, at a certain density for "normal" phone use distance will look poorer at any of the current screen sizes. What this article is trying to point out, is that the pixel density may be high enough for the problems with using pentile to not be noticeable.

    Please stop using monitors as an example, you're making terrible comparisons. 60% diagonal size loss on a monitor is much more extreme than the 16%, at most (4" to 4.65) on a phone.

    Hopefully the nexus' screen is good enough to make the pentile matrix not give poorer quality images as it does in some of the older phones
    Reply
  • zorxd - Tuesday, October 25, 2011 - link

    Of course I prefer a 36" monitor over a 20" one with the same resolution. If only I could afford it. That's why I chose a 22" monitor over a 20" with the same 1680x1050 resolution. Even if it was more expensive and has a lower pixel density.

    Going from 3.5 to 4.65" is a 33% increase in diagonal, as someone else said. So it's pretty big.
    Reply
  • a5cent - Friday, October 21, 2011 - link

    Hi zorxd, I think you've got this completely backwards. PPI is the much better metric because it is a measurement of quality! A higher PPI values mean the display will render text and graphics clearer. Since we all see differently, each of us will set our own minimal PPI value at which we no longer realize the on screen image is composed of individual pixels. This minimal PPI value is what all of us should be paying attention to when shopping for a device. If I'm shopping for a smaller device with a 3.7" screen, will I be happy with a 800x480 resolution. I have no way of knowing without knowing what minimal PPI I am comfortable with. As you have already mentioned, screen size on it's own is basically worthless information. Reply
  • zorxd - Monday, October 24, 2011 - link

    PPI isn't a measurement of quality.

    A 400 PPI screen is useless for a cell phone if it's under 2" or over ~6". Might be great as a watch or a tablet however.

    Let's take the original Desire. It has a resolution of 800x480. Its successor, the Desire HD, has the same resolution. However, the screen size went from 3.7 to 4.3". Is it a downgrade? Of course not. Some people will prefer a smaller phone, other a larger.
    But you can't argue that the smaller display is of better quality simply because of the PPI. What count is the resolution. Try to fit a web page on the 2" 400PPI display. It will look like crap because the resolution will be too low so you will have to scroll. However, a web page on a 1280x800 display looks just fine, no matter the size of the display, because web page are usually made for about that resolution.
    Both Desires will fit the same number of lines of a given web page on their screen because both of them have the same resolution. It won't look better on the smaller display, only smaller.

    A 960x540 4.3" display is of better quality than a 3.7" 800x480. Some people might accept the tradeof of getting a lower quality (lower resolution) display in order to get a more pocketable phone. There is nothing wrong with that. However, it would be stupid to argue that it would be an upgrade in quality.

    Before Apple, nobody ever talked about PPI of displays. We never cared before, because what counts is the resolution. If Apple could increase their display size to 3.7" without making the phone bigger, it would be a very nice upgrade even if it would reduce the PPI. Unfortunately their marketing department won over the reason so they can't do that in any near future.
    Reply
  • a5cent - Friday, October 21, 2011 - link

    Sorry, that last sentence should read:

    As you have already mentioned, screen RESOLUTION on it's own is basically worthless information.
    Reply
  • Rurou - Friday, October 21, 2011 - link

    Both the screen size and PPI have their use. But your example isn't exactly perfect either. If PPI is the most important thing for your photo viewing, then why don't you use 10" 720p screen?

    What I'm trying to say is that of course PPI is important, but screen size is equally important. You can't easily separate the PPI needs based on usage. Both are closely related. For me, as long as text looks crisp enough at their intended usage distance then I would choose the bigger screen even if it means sacrificing PPI. Of course for portable device (laptop, cellphones, etc) the size consideration is one of the priority when choosing the device. When I'm choosing my laptop, I have a target size. For example: when buying laptop, the first thing I choose is the size. I want a 12-13inch because the screen isn't too small and it's still portable enough. Sure, I can choose those 10" screen for extra PPI, but the screen would look too small. Anything bigger than 13" would be too big. Now, if there is a choice between 13" 720p vs 13" 1080p then of course I would choose the latter as long as the price is within my budget.

    Like the current iPhone screen situation. A lot of people wants bigger screen. They don't really care if the bigger screen would reduce the PPI, they just think that the current screen is too small.

    In Galaxy Nexus case, the screen (for me) is too big, thus I would hold it further from my face compared to when I hold my 3.8" phone... even then my 3.8" is crisp enough. So the fact that this one have a bigger screen (but still small enough) and HD res should be enough to compensate the fact that it is using PenTile layout.
    Reply
  • deV14nt - Friday, October 21, 2011 - link

    It also tends to be ignored that the screen is not just a screen for you to look at. It is a touchscreen. An input device. Certainly screens bigger than 3" are more usable than those that aren't (remember 2.8" Touch Pro?). Can we all agree on that?

    The upper bound on usability is where you can no longer use the phone one-handed to dial. I'd argue that the experience improves gradually all the way up to 4.3 or more, where it drops off sharply. Even bigger than that when they remove the hardware buttons like they did on this phone.

    Another one of many things that get ignored when all you choose to see is pixel density.
    Reply
  • zorxd - Monday, October 24, 2011 - link

    You can dial with one hand on any screen size. They only need to make the dialer app small enough so that all numbers stays within range. Reply
  • Gansan - Friday, October 21, 2011 - link

    If you have to hold the display further away, that's the same as making the display smaller. You could also say if the 3.5" display is too small, you could hold it a couple of inches closer. Reply
  • zorxd - Monday, October 24, 2011 - link

    That's not true because too close it's impossible to focus.

    For the same reason, people buy 60" TVs instead of buying 32" TVs and sitting just in front of it.
    Reply
  • name99 - Sunday, October 23, 2011 - link

    It depends what you do. If you read a lot on your phone, then high pixel density really does make the text crisper and that much easier to read.
    To take a realistic example - text on my iPad (under normal reading conditions appropriate for the device) is still at the point where jaggies can be seen on the diagonal edges of letters like X. Text on an iPhone4 is no longer at that point, those jaggies are not seen. This is precisely the reason why those individuals who primarily use their iPads as eReaders care so much about when iPad will get a high-res screen.

    Sub pixel numbers give one some insight into how much better a screen will be in this respect because nowadays everyone uses some sort of subpixel optimization for their text rendering. I don't know if the metric Anand uses in the second graph is perfect for this purpose, but it's probably acceptable. Everyone, when they see text on an iPhone4 screen, says "wow". I'm unaware of that sort of reaction from any of the current crop of AMOLED screens.

    Now, if you're mainly playing games or watching movies, I suspect it, to be honest, doesn't make THAT much difference. Certainly for me, movies were just fine at the resolution of my iPhone1, but I'm not obsessive about movies and I don't play games, so you should ignore my opinions on these points. Text is what I care about, and for text, well, you've seen my opinion.
    Reply
  • zorxd - Monday, October 24, 2011 - link

    By any standard the iPad is better for reading than the iPhone. You will fit more text on the display (because of the slightly higher resolution) and the bigger size will make it easier to read.
    The downside is that it doesn't fit in your pocket.

    So I don't get your point at all.
    Reply
  • tealchip - Wednesday, October 26, 2011 - link

    The reason we care about pixel density is to approximate paper ever more closely, resolution-wise. Higher pixel density means cleaner graphics, nicer typography, and less need for antialiasing.

    Pixel count matters too, but density is ultimately what yields better-looking displays. Maybe you need to be a designer to fully appreciate it, I dunno, but I like to think people notice sharper, crisper graphics without jaggies or antialiased fuzz.
    Reply
  • shorty lickens - Wednesday, December 21, 2011 - link

    Ummm I care about the pixel density, I care a lot.
    800x480 on the Titan looks freakin awful.
    800x480 on a 3.7 inch screen looks quite nice.
    It matters, a lot.

    Of course, 1280x720 on the Nexus looks freakin sweet and will hopefully raise the bar for expensive smartphones in the future.
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    zorxd said:

    'why do we even care about these charts?

    What we care is the total number of pixels. Not the density.

    You wouldn't want a 400 PPI 320x480 display on your cell phone.
    Just like you prefer that 23" 1920x1080 desktop display over a 300 PPI 1024x768 one.'

    The answer is that with Pentile you aren't getting full pixels. Each pixel is effectively Red, Blue and HALF a green (shared with the next pixel).

    So if you care about pixels, you care about Pentile.
    Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    We confirmed this a while ago at http://www.thegadgetlife.com.au, when the Galasy S II HD and Galaxy Note were first announced with similar pentile screens.

    While it is true that it should at least be comparable to the Galaxy S II screen, the compromise doesn't make sense except for marketing - why modify the perfectly good SGS II screen simply to be able to pass it off as an 'HD' screen?

    I don't quite understand how you calculated the subpixel per inch figures above but they appear to be wrong - check http://thegadgetlife.wordpress.com/2011/09/03/the-... for an explanation. The calculation sohuld be simple, the iPhone 4S and Samsung Galaxy S II have 3 subpixels per inch, so multiply by three for them, and by 2 for pentile screens.

    So the subpixel density is:
    iPhone 4S: 990
    Galaxy Nexus: 632
    Galaxy S II: 651
    Galaxy S/ Nexus S AMOLED: 466
    Droid Razr: 512

    I don't know how you made your calculations but they appear to be wrong. The Galaxy S II actually has a slightly higher sub pixel density.

    You also don't mention that the Droid Razr also uses a pentile RGBG screen (again check http://www.thegadgetlife.com.au for proof and details). The subpixel density for the Droid Razr is 256x2 i.e. a mere 512ppi.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    This is not the correct way to calculate subpixel density, as you cannot simply scale the effective pixel density by 2 or 3 in that manner. Recall that subpixel rendering layouts affect the density in horizontal space, not vertical space, eg there are three horizontal subpixels per logical pixel in RGB, two subpixels per logical pixel in RGBG or RBGW PenTile. Thus, the computation is a bit more involved than merely scaling.

    The computation in this case involves calculating the horizontal subpixel density and vertical density (which is the same as vertical resolution in this case), then calculating.

    To imply that there are 990 subpixels per inch in the iPhone 4/4S should immediately raise an eyebrow, as this implies that logical pixels are comprised of 3x3 subpixels, which is obviously not the case.

    So in reality, my computation is not incorrect, but you may want to revise your own blog post :P

    Also I've seen the photo in mention on droid-life and while it is immediately obvious their subpixel rendering layout is not RGB, it isn't clear to me what it actually is.

    -Brian
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    Brian, are you sure?

    Consider a screen which has 9 ppi:

    (here is one inch)
    xxx
    xxx
    xxx

    (3)

    Each of those pixels is made up of 3 pixels:

    (same one inch in sub pixels)
    xxx xxx xxx
    xxx xxx xxx
    xxx xxx xxx

    (27)

    Thus, you can simply multiply by 3.

    It's obvious really, PPI is a measure of pixels. 1 pixel = 3 subpixels. So to go from pixels per inch to sub pixels per inch, you multiply by 3. Thus the iPhone 4S has 990.
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    The (3) should read (9) Reply
  • PWRuser - Friday, October 21, 2011 - link

    "While it is true that it should at least be comparable to the Galaxy S II screen, the compromise doesn't make sense except for marketing - why modify the perfectly good SGS II screen simply to be able to pass it off as an 'HD' screen?"

    Hmm, I wonder why. Maybe because the HD resolution is a natural progression after ~3 years of WVGA? Or maybe because LG, Hitachi and others are about to introduce their HD displays? Or maybe because a ~200 pixel density doesn't cut it in 2012?
    Reply
  • lammers42 - Friday, October 21, 2011 - link

    The super amoled plus screens were consuming much more power than the older super amoled. I thought that the HD super amoled was supposed to correct the increased power consumption and keep the RGB subpixel layout, but I guess we wait until next year for that. Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    Still trying to make sense of how you calculate suboixel densities. As I understand simply multiplying the ppi by 2 or 3 for pentile and RGB stripe displays respectively should give you the subpixel density. Reply
  • Brian Klug - Friday, October 21, 2011 - link

    I explained this above in my original reply, but the trick is that there is only one vertical pixel per subpixel. This absolutely must be taken into account.

    Stated another way, if we're counting subpixels, one must take into account that RGB and RGBG/RGBW merely affect the horizontal subpixel rendering layout,

    eg: two "pixels" in both lands: RGB, RGB = ||| ||| and RG, BG = || ||

    There isn't a change in the vertical direction, thus you cannot simply scale the pixel density by some scalar. Instead, you must recompute the actual horizontal resolution (which is 2*horizontal resolution for RGBG or RGBW, and 3*horizontal resolution for RGB) and vertical resolution (which is unchanged), and then pipe those actual subpixel resolutions into your pixel density calculation, which is:

    sqrt((horizontal)^2+(vertical)^2)/(size in inches)

    Then you will arrive at the actual subpixel density in subpixels/inch.

    -Brian
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    Brian, you don't need to use any square root in your formula.

    The number of sub-pixels in an inch is simply:

    (horizontal sub pixels * vertical sub pixels ) / size in inches

    ie, the number of sub pixels / inch

    For some reason, you are calculating the diagonal length, in sub pixels.

    Look at the examples you gave of pixels (lets assume this is a 2 ppi display with a resolution of 2x1 and a screen size of 1 inch squared)

    RGB: ||| ||| - 2 pixels, 6 sub pixels (2 x 3)
    RG BG: || || - 2 pixels, 4 sub pixels (2 x 2)

    Plug your own formula into this, and you get:

    RGB: sqrt (( 6^2 + 1^2))/ 1

    or:

    RG BG: sqrt (( 4^2 + 1^2)) / 1

    This gives sqrt (37) for RGB which is 6.08. But there are 6. For the RG BG it gives sqrt (17) which is 4.12. But there are 4. So your formula is wrong. Note that what you've actually calculated is some bizarre form of Pythagoras which measures the length of the diagonal of the screen, in sub pixels. And then divides by the area of that screen, in inches. It's a meaningless number.
    Reply
  • sjael - Friday, October 21, 2011 - link

    Have to agree - I can't figure out how they got the second graph at all.

    In addition, it's not fair at all to compare RGBG PenTile to RGB Stripe solely on the number of sub-pixels employed. RGBG PenTile still has the same number of green sub-pixels, which are the most important to image detail. And it's not just that - PenTile doesn't even really draw to physical pixel coordinates like a standard display does.

    Even so, at its absolute worst (solid blue text on a solid red background?) the PenTile screen shouldn't really look any worse than a qHD display at the same size. At its best it will look far better.
    Reply
  • sjael - Friday, October 21, 2011 - link

    (And just before Brian gets on my ass about my first statement, yes, I actually looked at it in my head about three seconds after clicking Post and realised how he did it :P) Reply
  • Brian Klug - Friday, October 21, 2011 - link

    See my above two replies where I outline how the actual subpixel resolution of a display is not simply a scalar multiple of the effective pixel density.

    Also, explain how it is not fair to compare RGB PenTile to RGB based on the true subpixel density of the two displays? Most people complain that in either of the PenTile schemes, they can see the subpixels when shown solid colors. Thus the normal pixel density comparison as a function of effective resolution is not a fair means of comparison since it's predicated on there being three subpixels per pixel (and judged accordingly).

    True, PenTile does draw at the subpixel level, but the framebuffer that the PenTile driver is looking at effective resolution. I don't disagree that it won't look any worse than a qHD display, the point is that 720p PenTile shouldn't be nearly as distracting as other RGBG AMOLED panels we've seen before (eg Nexus One/Nexus S which are near the bottom).

    -Brian
    Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    Green is more related to the the perceived brightness than level of detail as I understand.

    For level of detail, if you have a one pixel width white line for example, there isn't much clever pentile rendering can do about it as individual pixels in the pentile layout simply can't show the colour white (or gray or really any colour that has no zero values in the RGB triplet)
    Reply
  • sjael - Friday, October 21, 2011 - link

    They use more green sub-pixels because visual acuity peaks in green. Thus having more green subpixels allows the display to have basically the same perceived resolution (as RGB) in any color containing green. They cut out half the blue and red ones because we see less detail in blue and red, so for most purpose we don't notice their absence.

    It's quite a bit more complicated than that, but that's the basic principle as I see it. (See what I did there? :P)
    Reply
  • grkhetan - Friday, October 21, 2011 - link

    When is iPhone 4S review going out? Reply
  • imaheadcase - Friday, October 21, 2011 - link

    Just read a regular iphone review. Reply
  • Brian Klug - Friday, October 21, 2011 - link

    When it's finished :P

    By the end of this week most likely, no later than Sunday.

    -Brian
    Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    Let me think about this, but I have a feeling you are wrong.

    There is an effective change in vertical resolution as well.

    for example, the SGS 2 with an RGB layout in subpixel terms is

    800*3 X 480 *3 NOT 800*3 X 480

    There is a change in vertical density as well, because each pixel contains subixels, not just vertical ones.

    Take a 2x2 one square inch screen.

    RGB layout:

    RGB RGB

    RGB RGB

    that's 4ppi or 12 subpixels per inch (call it spi)

    RGBG pentile layout:

    RG BG
    RG BG

    that's 4ppi or 8 spi

    Unless my understanding of what ppi means is wrong, but I don't think it is.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    You are partially right and partially wrong, and you're still missing the point that there is not a vertical change in resolution between the two, and that your scaling of PPI by 2 or 3 does not represent the subpixel density. The problem isn't in your diagram or understanding, it's in the fact that you turn around and just scale PPI by some scalar instead of actually doing the math yourself, and no, it is not 800 * 3 x 480 * 3.

    Again, what you're doing when you're setting up 800 * 3 x 480 * 3 is

    RGB RGB
    RGB RGB
    RGB RGB

    RGB RGB
    RGB RGB
    RGB RGB

    for one

    | |

    | |

    If you look at the actual subpixel, you'll notice that it is in fact

    RGB RGB
    RGB RGB

    again, the vertical resolution is unchanged between RGBG and RGB. The two subpixel rendering schemes are 2, 3 respectively in horizontal space.

    Another example that might help this click, look at nuovoyance's own diagram: http://www.nouvoyance.com/images/rgb-v-rgbw-circle... then count the number of horizontal pixels in both. Note that the height of each subpixel is unchanged.

    -Brian
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    Again, Brian, look at this example.

    For RGB - you drew 12 sub pixels for a 2 x 2 pixel display.

    Which is 3x the number of pixels.

    You are correct that only the horizontal number if amended, which is why we only multiply by 3, and not by 9 (if a pixel was made up of a 3 x 3 array we would have 9 sub pixels for every pixel).

    But you haven't multiplied by 3 in your table (for the iPhone). You should have.
    Reply
  • smeee - Friday, October 21, 2011 - link

    And this is why I read AnandTech, not "thegadgetlife.com.au"... Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    No kidding. Even at 1 am, AT is still sharper than the Aussies. Go Team Klug! Reply
  • Gasaraki88 - Friday, October 21, 2011 - link

    Yeah, especially when he kept arguing that he's right. Reply
  • miketlo - Friday, October 21, 2011 - link

    If it was 800*3 x 480*3 you would have too many sub pixels. The total number of subpixels is 800*480*3 since each pixel has 3 subpixels.

    By your counting of 800*3 x 480*3 on the same scale 2x2 one inch square you would have this layout:

    RGB RGB
    RGB RGB
    RGB RGB

    RGB RGB
    RGB RGB
    RGB RGB

    Since you defined there to be '3 vertical' and '3 horizontal' subpixels per pixel in your definition and the way you counted.
    Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    Never mind, I am stupid sometimes, my bad folks :D Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    Just calculate the area and use that.

    On a 100x100 grid, you have 10,000 total pixels (i.e. 100*100).

    So, there's 30,000 total subpixels (i.e. 10,000*3) on our grid.

    So how did we calculate that? (hint: use the parenthesized calculations)

    30,000 = 3*10,000 = 3*(100*100)

    So, we can use that result to explain the Klugian Method.

    3*(100*100) = 3*100*100 = (3*100)*100 = (3*100)x100

    If you don't have a strong understanding of area, it can be hard to immediately jump to Klug's explanation.

    Then again, if you don't have a strong understanding of area, why are you managing a tech blog? O.o
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    Your definition is correct.

    But it doesn't fit with what Brian actually did.

    He took a ppi of 329.65, and worked out a sub pixel density of 613.32.

    It should have been 988.95 (329.65*3)
    Reply
  • catnaps - Friday, October 21, 2011 - link

    You're calculating based on an incorrect assumption. Each pixel is made up of 3 sub-pixels, not 9. Why 9? Because you're including vertical resolution in your calculation, which is wrong because it's not increased- horizontal is.

    It's really quite a simple concept to get your head around. Back away from your understanding a little and challenge the observations you're making.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    One thing I will say is that this was intended to be a quick means of showing where the display should fall.

    I believe a clearer way of showing this difference would be to actually compute a weighted sum of the displays, eg

    (720*1280/2)*2/3 + (720*1280)*1/3 = 614,400 pixels in RGBG land

    as opposed to

    (1280*720) = 921,600

    and then

    614400/921600 = 2/3

    which shows exactly the 'RGBG uses 2/3rds the pixels of RGB' line. Then you can do sqrt(2/3) = 0.8164, and then scale each dimension accordingly, eg 0.8164 * 1280 = 1045 0.8164 * 720 = 588, so then the display is "effectively" 1045 x 588.

    Then work backwards, to sqrt(1045^2 + 588^2)/4.65 = 257.8 effective subpixels per inch.

    The problem is that tackling this either has to happen in subpixels/inch, or pixels/inch. And mixing one up in place of the other makes results that aren't comparable.

    -Brian
    Reply
  • doobydoo - Saturday, March 17, 2012 - link

    If you have a device with a resolution of 1045 x 588, you can't work out the DPI in either pixels or subpixels, using any formula.

    As I said, you're calculating some weird diagonal in sub pixels, and then dividing it by the screen diagonal.

    In fact, knowing the following information:

    Horizontal Resolution
    Vertical Resolution
    Diagonal Screen Size

    Is simply not enough to work out the DPI. This is because the diagonal doesn't give you the size of the screen in terms of area. The same length of diagonal can actually represent a smaller or larger surface area. You have to know either the width, the height, or the ratio, before you can calculate the DPI (or the sub pixels per inch).
    Reply
  • Taracta - Friday, October 21, 2011 - link

    A 2x2 one square inch screen with RGB layout:

    RGB RGB

    RGB RGB

    Would give you 2ppi or 6spi NOT 4ppi or 12spi, this is because for every inch across, horizontal, there are 2 pixels and for every inch up or down, vertical, there are two pixels. You will notice that sub pixels are totally dependent on the whole pixel and you have to could the sub pixels in each pixel and not just the leading sub pixel.

    The 2x2 one square inch layout of RGBG pentile arrangement it is a little more complicated:

    RG(B)GR

    BG(R)GB

    Would give you 2ppi or 5spi, the center sub pixel is shared by both pixels and it is never the G(reen) in the current usage.

    I believe that you are correct in stating that to calculate spi you would x3 for RGB, for both vertical and horizontal, and for the pentile RGBG it is approximately x2, FOR THE HORIZONTAL MEASUREMENT and x3 for the VERTICAL MEASUREMENT.
    Reply
  • Chinpokomon - Friday, October 21, 2011 - link

    Maybe it doesn't matter for the purpose of this discussion, but the penile layout is actually more like:

    RG BG
    BG RG
    Reply
  • sjael - Friday, October 21, 2011 - link

    "Also, explain how it is not fair to compare RGB PenTile to RGB based on the true subpixel density of the two displays? "

    I didn't say density though, I said just the number. What I mean is that PenTile has gotten a lot of bad press because now everyone knows* that PenTile has "two subpixels per pixel instead of three" and is thus always horribly inferior. However in most cases (that aren't *very* small lines on contrasting backgrounds :P) there's no noticeable difference (unless you have that particular mutation,) and especially in this case it allows a much higher resolution than was possible with RGB Stripe.

    Even in the Galaxy S, I assume the relatively low pixel density was down to yields at the time, and that density allowed them to put into phones a stunning OLED screen that nobody really had a problem with because no one knew there was something to criticize yet.

    I guess I'm just tired of reading all day about how Samsung/Google "cheaped out" on the screen. From every angle I've seen, it appears to be right up near the limit of Samsung's current AMOLED printing technology.

    I do agree on 315ppi PenTile probably not being noticeable though - I would be very surprised if any of PenTile's critics could tell it apart from RGB stripe at that density.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    I think we're actually in agreement then, PenTile actually has a lot of promise as far as different subpixel rendering layouts go, the problem is that the two most popular implementations thus far have had subpixels that were larger than visual acuity (well, at least mine :P), and thus don't exceed it enough to deliver the promise of - same effective resolution with fewer subpixels.

    I believe that in this case, with 720p and 4.65", it might finally be good enough to emulate that higher resolution RGB stripe. I definitely agree that Samsung/Google didn't cheap out here, if anything this should illustrate that the subpixels in RGBG 720p PenTile should be smaller than those in the RGB SGS2, and likely are made with a correspondingly smaller geometry on their mask - eg more advanced. I guess that's my point with the line where I state it's above SGS2, it really should be better.

    -Brian
    Reply
  • miketlo - Friday, October 21, 2011 - link

    You beat me to the exact same explanation while I was typing mine up. +1 :) Reply
  • thegadgetlife.com.au - Friday, October 21, 2011 - link

    Well instead of talking about horizontal and vertical resolutions, can you tell me what's wrong with the following reasoning:

    A 100×100 pixel image that is printed in a 1-inch square has a resolution of 100 pixels per inch (PPI). (from wikipedia)

    Since each pixel in RGB has 3 subpixels, for RGB layouts you you can write this as:

    A 100×100 pixel (OR 300x300 subpixel) image that is printed in a 1-inch square has a resolution of 100 pixels per inch (PPI) OR 300 spi. (from wikipedia)

    In the above example, there are 10,000 pixels per square inch. And all 10,000 pixels have 3 sub pixels. So the subpixel count is 100*3 X 100*3, NOT 100*3 X 100.

    If you represent a pixel as X, the layout of the 2x2 one square inch display example I gave above is:

    X X
    X X

    if you replace each pixel by its RGB subpixels, it becomes

    RGB RGB
    RGB RGB

    or for RGBG, it becomes:

    RG BG
    RG BG

    I may be completely wrong here, but still want to figure out how.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    The problem is that your definition and layout is correct, but you aren't actually doing the math per-se.

    Eg:

    sqrt(6^2 + 2^2)/1 = 6.32

    sqrt(4^2 + 2^2)/1 = 4.47

    In both cases, there is one horizontal subpixel per logical pixel. The problem is that the square in there throws off that scalar scaling.

    It's a case of, the problem is being setup properly, but not mapped to math properly, I think.

    -Brian
    Reply
  • InternetGeek - Friday, October 21, 2011 - link

    I'm not an expert, but I think that multiplying by 3 the dimensions just because an RGB pixel has 3 subpixels is wrong. I believe it depends on the arrangement of the 3 subpixels.

    So if the arrangement is one pixel as a kind of pivot with one on top and one on the side; you would multiply by two to obtain the number of vertical subpixels and by one to obtain the number of horizontal subpixels.

    In PenTile, the subpixels are arranged horizontally, that is only one vertically and 3 horizontally. So vertically the number is not changed, horizontally the number is multiplied by 3.
    Reply
  • dukepeter - Friday, October 21, 2011 - link

    "In the above example, there are 10,000 pixels per square inch. And all 10,000 pixels have 3 sub pixels. So the subpixel count is 100*3 X 100*3, NOT 100*3 X 100."

    actually, IT IS 3*100x100. each RGB pixel is composed of 3 sub-pixels, not 9.
    100*3x100*3=90,000, not 30,000 as it should be...
    Reply
  • sjael - Friday, October 21, 2011 - link

    Exactly, case closed.

    And I really do need to compliment you guys - you are the only tech site I've ever used where the writers actually engage in dialogue with commenters. For that reason Anandtech is the only site I actually take the time to comment on. Big thumbs up.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    That's part of the job :)

    You guys keep me on my toes too, which is the awesome part, because sometimes we really do need to be corrected or set straight. I'm not above making mistakes at all, especially trying to calculate subpixel density at 1 AM PST, hahaha

    -Brian
    Reply
  • Kurian..Of Borg - Friday, October 21, 2011 - link

    Scam screen is scam. Reply
  • jalexoid - Friday, October 21, 2011 - link

    "Galaxy Nexus display does in fact use RGBG PenTile, like all the Nexus devices to date"
    Nexus devices usually come in 2 flavours - AMOLED and sLCD. sLCD uses RGB
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    Both devices, the Nexus One and Nexus S, were introduced with AMOLED screens, the S featuring a SuperAMOLED. Both were eventually sold with sLCD's but his statement still stands, all Nexus devices used RGBG Pentile. Reply
  • tdawg - Friday, October 21, 2011 - link

    I'm glad to hear that the Nexus One was a pentile screen. I was getting worried that I was going to have to skip the Galaxy Nexus due to a subpar screen, but I was never dissatisfied with the screen on my N1 so I can't imagine I'll be dissatisfied with the Galaxy Nexus screen. Reply
  • luimana - Friday, October 21, 2011 - link

    Thanks for your article Brian and Jason.
    Just a very small note: iphone 4 ppi is not 329 but 326. The screen infact is not 3.50 inches wide but 3.54

    Luigi
    Reply
  • orion2001 - Friday, October 21, 2011 - link

    I'm still rocking a rooted OG Droid, and I've been looking forward to the Nexus and the RAZR as potential upgrades. However, the second chart in this article is just so damn disappointing. Why have smartphone manufacturers gone backwards with the displays they are using on their phones? I really like the screen on the OG droid, but I would at the very least expect something comparable, if not better out of the latest tech. The higher pixel (sub-pixel) density really makes a big difference when you are trying to read small text on your phone.

    The Droid does a pretty good job at that, but it is just completely unacceptable for the latest android phones to be going backwards in this department. I've played around with some pentile display phones (I was considering the Droid 3 for upgrade a while back), but I just couldn't stand the screens. I could actually see each pixel quite clearly, and the same webpage looked way clearer on my Droid than the newer phones.

    And while AMOLED is nice and all, no one talks about how overly saturated and blown out the colors look on most of these displays. As someone who is into photography and has spent time calibrating his LCD monitors at home, the color accuracy on most AMOLED screens seems extremely poor.

    I can't believe that over a year after the whole Retina display hype, android phone manufacturers aren't even anywhere close to playing catchup with Apple on the display side of things. For all the crap Apple got for an incremental upgrade with the 4S, it still seems to beat the latest and greatest android phones on the display and GPU front. I can't stand IOS, and I love android, but I'm just really disappointed with what I'm seeing with the latest and greatest of the new android phones.
    Reply
  • dagamer34 - Friday, October 21, 2011 - link

    The advancement here is building an OS that supports 1280x720, which actually takes a lot of work to do to look good. The HTC Vigor has a 720p phone thats going to be released soon with a 720p display that's sLCD with an RGB sub pixel layout that'll look far better running ICS than Gingerbread because of the HD screen support. Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    I too share our love for the OG Droid and have only recently retired mine. There's no doubt that we would like to see more advanced displays, the question is cost. The technology to develop IPS displays with high pixel densities and good color characteristic exist, but gets more and more expensive the larger the display. The technology to develop AMOLED displays with high pixel densities and good color characteristics exist . . . with PenTile. Without Pentile there's a matter of mass production readiness. If you want proof that it's still not economically viable, look at LG who introduced their 720p HD display, and then quit the small OLED business completely.
    In two year we've gone from small (<3.7") LCD's of the TN variety being the only game in town in phones, to much larger (>4.0") LCD's of IPS, TN, sLCD and AMOLED varieties. We are swimming in great screen technology that provides not just great images, but options. Is RGB so important to you? Then go get an SGSII, go get an Atrix 2. If you don't mind small screens, go get an iPhone, or even (as a small upgrade to your phone) a used Droid 2. We've got options, we should enjoy them!

    Jason
    Reply
  • orion2001 - Friday, October 21, 2011 - link

    Jason, I agree with you to some extent, but if the Droid could afford the screen it had...and with all the advancement in display tech in the last 2 years, I just don't see why they are getting away with putting more inferior displays into the phone.

    And it is even more painful to see, that I can get the Iphone 4S (which has the more expensive components) for $50 less than the Droid Bionic on Verizon. Most likely, the Nexus and RAZR will be even more expensive. If I was your average person who wanted to buy his first smartphone, there is no way I'd pay more for phones that are not close to proven compared to the iPhone. I've accepted that with tablets, it is harder for manufacturers to outcompete Apple on pricing due to the money Apple makes with their appstore. But at least when purchasing a phone, I always assumed that carrier subsidies would price Android and Apple phones equivalently. When the former is cheaper by $100 in comparison to the latest Android phone, you can start to question whether it is worth it.

    Make no mistake, I can't stand Apple products, and I will buy an Android phone. But if I was recommending a phone to a friend who wasn't ultra techie, I'd recommend the iPhone 4S for a better phone at a cheaper price.

    As for getting the SGSII, I really would...if it ever comes to Verizon. That is the only phone that I have really liked and been impressed by, and it is a shame that even a year down the road, I haven't seen any phone that has taken the next step forward beyond the SGSII.
    Reply
  • marc1000 - Friday, October 21, 2011 - link

    Brian, what about the N9 display? how does it stack up in this comparison?

    i've read somewhere that it uses a pentile grid too, but it is smaller than those 4.5" giants.... it's said to have a polarizer, wich is something i never saw on other phones, but that pentile "karma" made me kinda sad.

    (and i've not had access to the phone yet to check if it actually looks good).

    thanks in advance!
    Reply
  • Jussi7 - Friday, October 21, 2011 - link

    I can confirm the N9 screen uses pentile grid. For me the screen was the biggest letdown in the phone, text looked fuzzy and I could see the patterns in color fills that should have been solid. Also the brightness was much less than I expected. With sensory assessment the text was a tad sharper than on my iPhone 3GS, iPhone 4 was from another planet.

    If I did the math[1] correctly N9 screen would be at 429 sub pixels per inch.

    [1] sqrt((480*3)^2 + 854^2) / 3.9
    Reply
  • marc1000 - Friday, October 21, 2011 - link

    thank you Jussi. i was looking for a confirmation if the text or solid colors really looks fuzzy on the N9. this was also a big letdown to me. i really liked the way meego works and the whole "button-less" experience... Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    Ahh! You forgot, that's (480*2) not 3. You don't live in Philly do you? Disappointment or not, I'd love to play with an N9 for a little while. Reply
  • Jussi7 - Friday, October 21, 2011 - link

    Yeah, I messed up the math, sorry about that. The new, hopefully better number is 329 sub pixels per inch.

    But the number does not really matter, the screen was disappointing. On the other hand the scrolling and transitions were very smooth. I mostly enjoyed the swipe based UI concept, although I kept doing some accidental swipes every now and then. I played with it just a few hours though, easily the best Nokia so far.

    I do support the Eagles, and hate the Patriots, but no, I don't live in Philly. I currently reside in Helsinki, Finland. :)
    Reply
  • luimana - Friday, October 21, 2011 - link

    Hi Brian,

    I've followed the debate on how should the spi density be determined. I agree with your logic, but I come up with different numbers:

    e.g iphone: 326ppi -> 3 sub pixel per pixel (only on one dimension) -> 326*sqrt(3)= 564,5spi

    You used 329ppi in your graphs so you should obtain 569spi... How did you get 613?

    Thanks
    Luigi
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    The calculation is laid out in the discussions above but the equation we're using is explained nicely here (http://en.wikipedia.org/wiki/Pixel_density#Calcula... Essentially we're finding the diagonal resolution and then dividing that by the diagonal screen size. This is the mathematically right way to calculate pixel density. To calculate subpixel density you multiply the horizontal resolution (when in portrait!!) by the number of subpixels per pixel (2 for RGBG and three for RGB). Then calculate it out. For the iPhone 4:

    640 horizontal pixels X 3 subpixels/pixel = 1920 horizontal subpixels

    diagonal subpixel resolution = SQRT (960^2 + 1920^2) = 2146.6 diagonal subpixels

    subpixel density = diagonal subpixel resolution / diagonal screen size = 2146.6 / 3.5 = 613.3 ppi

    Now, we did err in not using the finer screen size measurement of the iPhone that's used sometimes (3.54") but you'll forgive us ignoring those four hundredths of an inch, I'm sure.
    Reply
  • marc1000 - Friday, October 21, 2011 - link

    yeah i forgive you about those four hundredths of an inch of the iphone. and i used your step-by-step to calculate the pixel density based on the specs of Nokia N9.

    the number i came to: 329,45

    wich seems pretty low in the graph... right next to HTC Droid. Is Droid's screen bad? or good?

    my calculations:
    480 horizontal pixels x 2 subpixels/pixel = 960 horizontal subpixels
    diagonal subpixel resolution = SQRT (854^2 + 960^2) = 1284,87 diagonal subpixels
    subpixel density = diagonal subpixel resolution / diagonal screen size = 1284,87 / 3,9 = 329,45 ppi

    am i right?
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    The N9 is a great phone, but is compromised in it screen by being AMOLED, Pentile, and either a little too big or a little too small, depending on how you look at it. A little bigger and we could forgive it it's low resolution for being easier to type on (though I understand it's actually got a dreamy keyboard). A little smaller and it's density would be higher and it would satisfy a bit more. Reply
  • luimana - Friday, October 21, 2011 - link

    Of course I forgive you (and hope you'll forgive me if I was so pedantic ;-) )
    Thanks for the clear explanation and the link.

    Actually I happen to use a different formula which was equivalent to determine the ppi but actually not correct to determine the spi.
    I'm always happy to learn something new, that's why I come and read Anandtech!
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    Glad you could join us! Always appreciate the comments! Reply
  • luimana - Saturday, October 22, 2011 - link

    Thanks Jason,

    actually I've been following you since 1999 but I tend to be a lurker (I know I'm a bad guy!)... :-)
    I subscribed again because I forgot my old nick and password!

    Ciao!
    Luigi
    Reply
  • ajp_anton - Friday, October 21, 2011 - link

    I think your definition of subpixel density is flawed.

    First you have to think about what it means when you have something per a 1-dimensional axis, when the density is different in the two directions. The way you define it depends on the actual ratio of the screen, which i think is wrong. With your definition, you can make the subpixel density greater or lower if you simply cut the screen into smaller pieces without altering the pixels themselves.

    For example, you're assuming that the pentile subpixels are arranged horizontally (assuming a standing 720x1280 resolution).
    sqrt( (720*2)^2 + 1280^2)/4.65 = 414.33 subpel/inch.
    Now, if you cut the same screen into four slices of 720x320 pixels each, the density suddenly increases to
    sqrt( (720*2)^2 + 320^2)/2.49 = 592.42 subpel/inch (yes, the diagonal becomes 2.49 inches).

    A better definition would be to simply not assume any direction in those subpixels, so if there are 2 subpels/pel, then multiply the pixel density by sqrt(2), and by sqrt(3) with RGB pixels.
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    The flaw there is that it's not an assumption to say that the subpixels have an orientation, when you zoom in real close you can see the subpixels, and their orientation (as in with magnification). This would be the equivalent then of creating a 1440 x 1280 display (720*2 x 1280). And were you to make such a display you wouldn't determine it's pixel density by multiplying the pixel density of a less dense display by a somewhat arbitrary square root.

    I'm not sure about the results in your sliced example, can you show me how you got to 2.49 for the new diagonal resolution?
    Reply
  • ajp_anton - Friday, October 21, 2011 - link

    Yeah I know that they do have an orientation, but you're giving it too much credit.
    Imagine a very long screen, with the subpels oriented in the "long direction". The direction of the diagonal is almost the same as the direction that has the most subpixels, resulting in a higher subpixel density, even though the number of subpels would be exactly the same if they were oriented in the other direction.

    Or mathematically,
    The longer the screen is in the direction you have the subpels, the more influence that multiplication by 2 is going to affect the result.
    sqrt( (w*2)^2 + h^2)
    If w>>h, then the subpel density approaches 2 times the pixel density, and if w<<h, then the subpel density approaches the pixel density. Depending on the ratio between width and height, the subpel density will be somewhere inbetween.

    What you should do is simply calculate the number of subpels per area, and take the square root of that to get an "average" number per inch where their orientation doesn't matter.

    Diagonal in my sliced example:
    sqrt( 1280^2 + 720^2 ) / 4.65 = 315.83 pel/in
    720 / [ppi] = width = 2.28 in
    320 / [ppi] = height = 1.01 in
    diagonal = sqrt(w^2 + h^2) = 2.49 in
    Reply
  • luimana - Friday, October 21, 2011 - link

    That's interesting. You are right.

    >What you should do is simply calculate the number of subpels per area, >and take the square root of that to get an "average" number per inch where >their orientation

    I used to calculate it in this way. The results were the same for ppi but not for spi. Then Brian point me to that formula in wikipedia and I thought I was wrong. But now you made me understand that really the right formula is the area+sqrt.
    Reply
  • Brian Klug - Friday, October 21, 2011 - link

    This is actually the problem I struggled with when first starting down this road - the aspect ratio of the subpixels themselves becomes a factor, and our diagonal based calculations are affected by the aspect ratio of the display in that arrangement. The original pixel density calculation presupposes square pixels (which is more or less the case) but becomes a bit flawed when moving to something not square. In this case, I agree a better way is to calculate and average, but even then I feel like there's a better way of representing this.

    -Brian
    Reply
  • tipoo - Friday, October 21, 2011 - link

    I didn't even realize my Nexus S has a Pentile display as well until I looked it up just now. I don't see the problem, really. Everything looks gorgeous, and the Galaxy Nexus by all indications will be even more eyegasmic.

    I think what the problem is that the first RGBW PenTile Matrix LCD screens were pretty bad and the whole blogosphere crapped on them, even while the OLED versions of these screens are much better.
    Reply
  • StevenN - Friday, October 21, 2011 - link

    It is much like the Foveon VS the Bayer layout discussion. When you are talking about high contrast edges (text for example) pentile is at a severe disadvantage and it stands out like a sore thumb when compared to RGB based pixel layouts of even lower resolution.

    And yes, I have seen the Nexus displays and text is actually painful (as in headaches) for me to read on it as a result of the pentile display. Pentile has a bad rap for a very good reason.

    For images and video, pentile looks good.
    Reply
  • krumme - Friday, October 21, 2011 - link

    I am an old conservative man that have no television because, i think the old tube was better. Therefore i have a projector with DLP3 TI engine. Good black, and fairly accurate colours.

    Now, i understand the quality of the IP4 screen, and appriciate its sharpness. Its important and clearly visiable under usage.

    I dont know this amoled tech, but it clearly have a huge advantage for the pictures to all the horific quality all other LCD i have seen including ip4. Its gives far more accurate colours, and fantastic blacks.

    Now if Apple did have Amoled, we would be talking deep blacks and colour accuracy all the time with charts of this all over.

    There is an unfortunate sideeffect of the big brands to define, what should be in focus. But ofcourse thats the effect of the brand, and the value of it, but it should always be mentioned when discussing how much glass, steel and subpixel density is the right balance.
    Reply
  • ltcommanderdata - Friday, October 21, 2011 - link

    If I'm not mistaken, AMOLED is well known to have oversaturated and exaggerated colors. If you want color accuracy you'd go with an IPS screen like the iPhone 4 uses. IPS computer monitors are what are used in the photography and other creative industries. Reply
  • Drewdog343 - Friday, October 21, 2011 - link

    SAMOLED has it's advantages, color accuracy is not one of them.. Reply
  • krumme - Friday, October 21, 2011 - link

    You are right, sorry wrong formulation. They look a bit oversaturated to me too, but still gives the punch that is needed for a small screen imho. Its like colours have difficult escaping the small screens especially if there is light around. I dont know if others have the same impression. Reply
  • FATCamaro - Friday, October 21, 2011 - link

    What about the fact that there is a much higher inter pixel distances in pentile. It's almost as if the pixels are smaller so there is more "black" in between the pixels. This makes it pretty bad as well. Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    It shouldn't matter unless pixel brightness varies based on pixel size. Reply
  • VTEChump - Friday, October 21, 2011 - link

    Can someone calculate the subpixels per inch of the upcoming Galaxy Note? I'm too lazy to figure out the calculation:). The Note is a Super AMOLED display, 5.3", 1280x800 resolution. Reply
  • ImSpartacus - Friday, October 21, 2011 - link

    And since modern American education doesn't provide citizens with a solid understanding of fractions and their important (but seemingly trivial) applications in manipulating unlike units of measurement, people still struggle with this stuff. Reply
  • GuinnessKMF - Friday, October 21, 2011 - link

    Unfortunately your calculation is ignoring significant digits. For example the iPhone you calculated as 3.5" I'm assuming to get 329, but if you use 9cm you get 326, which is the stat provided by apple, which makes me assume that the screen is actually closer to 3.54". This is a minor difference, but you're displaying 5 significant digits on your graph, when you only have 2 significant digits in your source. Reply
  • JunkyJoe - Friday, October 21, 2011 - link

    The new Galaxy S II HD, to be released along side the Galaxy S II LTE in Korea, also lacks the "Plus" RGB stripe. Meanwhile, the non HD Galaxy S II LTE retains the Plus with a 4.5-inch WVGA (800×480)

    http://www.samsungmobilepress.com/2011/09/28/Samsu...

    Is the RGB stripe something Samsung is unwilling to implement across the board with any "HD" phone, not just the Nexus?
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    It's expensive. It'll happen in about a year. But for now this is pretty great. Really. Reply
  • Taracta - Friday, October 21, 2011 - link

    720 remains 720 it does not decrease in the vertical. Reply
  • Brian Klug - Friday, October 21, 2011 - link

    Oh very good point! I made the same mistake that spawned all of this in my attempt to approach the problem from another direction, haha

    -Brian
    Reply
  • schriss - Friday, October 21, 2011 - link

    Forgetting the numbers, can you compare opening a complex website on Galaxy S II and Galaxy Nexus? Like, despite all these subpixels talk (sounds like fake HD to me), would a browser still show much more content on Galaxy Nexus than Galaxy S II? I mean resolution is higher, so I assume yes. I have seen few Galaxy Note hands-on videos but don't remember anyone opening a browser and comparing what they see with other phones.
    What I need from Galaxy Nexus is to display more content in the browser, almost like on a laptop, with no need to zoom in as much as I would have to do on Galaxy S II, my current phone.
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    I wish I could say it was this simple but the phone's software can manipulate the render to better conform to the screen and that manipulation can be different between two phones.
    What I can say is that if you could pull up the same browser with the same settings on each phone it should show more of the website on the Nexus than on the S II. Cheers.
    Reply
  • schriss - Friday, October 21, 2011 - link

    Thanks, that's what I hope for. I will be using the same browser: Firefox Nightly builds. Too bad none of the hands on videos tried to show off the browser and how much of a website fits the screen and how is it readable.
    I wonder how clear the text is going to be.
    Reply
  • JasonInofuentes - Friday, October 21, 2011 - link

    And so to fill you all in, all of this started because Brian and I both were independently wrapping our heads around how best to gauge these displays in a data driven way. Obviously it's becoming more complex by the moment. Sigh. Reply
  • WaltFrench - Friday, October 21, 2011 - link

    How about we compare the number of sub-pixels per screen? That gets away from so much confusion.

    If you hold a 4" diagonal device at a distance of 8" from your eyes, each pixel will take up the same visual space (“subtend the same visual angle”) as a 5" diagonal gadget at 10" (if both screens have the same number of pixels). Maybe you want to compare holding both at 9", but that's obviously sub-optimal for both, in terms of blockier pixels or smaller-appearing screen, take your poison. We over-40 types might prefer bigger screens that we could hold farther away for presbyopia, but I actually enjoy my iPhone up real close because I like just a bit of edge from seeing some pixel detail.

    It's obvious that the PenTile layout works better for some things than others, but that's also true for the RGB stripe layouts. Our alphabets emphasize long vertical strokes, so vertical detail isn't as important as horizontal in distinguishing, say “darn” from “dam.” Since phones are used both portrait and landscape, you need very fine resolution to make small text legible in either mode.

    PenTile isn't as susceptible to H vs V but it is much MORE susceptible to fuzziness (eyestrain & slower reading) when you have common arrangements such as blue text on white screens, or the worst, Red on Blue (since there's the worst likelihood of being able to choose between red and blue when the driver wants to). How often do these cases come up? Well, in photos, it's common to have detail in red and blue, so PenTile is probably fuzziest for photos.

    While not particularly worse for B/W text. With optimized drivers that I have yet to see — actually, the opposite is what Android/Sammy seem content with — PenTile could deliver very sharp B/W text.

    Finally, a quibble: in dividing the iPhone's pixels per inch by the sub-pixels per inch, I *don't* get the square root of 3, as I think I should. How were the figures derived?
    Reply
  • WaltFrench - Friday, October 21, 2011 - link

    From DisplayMate.Com:

    “Most reviews of displays in Smartphones and Tablets are filled with imprecise and incorrect statements and analysis from reviewers that don't understand what the display specs actually mean, don't understand how to objectively evaluate display technology, and haven't performed any lab measurements to verify or quantify their conclusions. If any of this matters to you then read these expert articles...”

    This post, and virtually ALL the comments, validate DisplayMate's claims.
    Reply
  • ydgmdlu - Friday, October 21, 2011 - link

    The Atrix 2, which has a qHD Super AMOLED display, is shown in the first chart to have lower pixel density than the Galaxy Nexus. But in the second chart, it is shown to have greater sub-pixel density than the Galaxy Nexus. This is only possible if the Atrix 2 has an RGB-stripe panel. However, other places have reported that it also uses PenTile. Since Samsung controls Super AMOLED technology, for another vendor's device to boast a superior version of Super AMOLED is unthinkable. The Galaxy Nexus's display is already more advanced that the Galaxy S II's because the sub-pixels are so much smaller. I have to believe that the reason why the Galaxy Nexus has a PenTile display is simply because RGB at 315 ppi is not yet possible for AMOLED. Reply
  • ydgmdlu - Friday, October 21, 2011 - link

    Oops, my bad. I confused the Atrix 2 with the Droid RAZR. I stand corrected; the second chart is accurate. Reply
  • ajp_anton - Friday, October 21, 2011 - link

    What would be a better way?

    You basically have these choises:
    1. Worst case scenario, assume the orientation is along the shorter side.
    2. Best case scenario, assume the orientation is along the longer side.
    3. No orientation, evenly distributed.
    4. Other kinds of "averages" of 1 and 2 (e.g geometric or arithmetic means)

    Benefits of #3:
    - It is the only one that is unaffected by the screen ratio and how you cut it. It is constant and well defined for every size and shape of the same panel.
    - You only need one single element (pixel) to determine the density, you only have to know what kind of panel is used, not its size.
    - You don't have to know the subpixel orientation. In the future it might even become diagonal (two triangular subpixels per pixel, or something else in RGB pixels), in which case #3 is the only one that still works without modification.
    - If subpixel density becomes a popular marketing number in the future, #3 is the only one that can't be cheated by orienting the pixels in a way that has no benefit, but gives a bigger value.
    Reply
  • JesusR - Friday, October 21, 2011 - link

    I always see the iPhone as the one with more pixel density, but how about the LG LU6200, with 720×1280 in 4.3" (i.e. 342 ppi)? Or the several Sharp phones with 327 ppi? Or even the Toshiba Portege G900, with 313 ppi in 3'', whcih was anounced in February 2007, 3 years before Apple presented its Retina Display? I think there are several phones with similar pixel density. Reply
  • Chinpokomon - Friday, October 21, 2011 - link

    Using the same calculations in the article for effective pixel density, I got the same answers in the first graph (updated to 3.54" for the 4S).

    I then calculated the actual pixel density rather than subpixels, and got:

    iPhone 4S - 325.93
    Galaxy Nexus - 293.99
    Nexus One - 232.87
    Galaxy S II - 218.49
    Galaxy/Nexus S - 215.41

    This uses a (2/3) scaler for the horizontal pixel resolution, thereby ignoring that we no longer have square pixels in the calculation. I then calculated the actual diagonal resolution and found the actual PPI. Too me, this is a better way to compare both the effective and actual PPI of these different displays.

    The "Retina display" display as explained by Chris Brandrick in this (http://www.pcworld.com/article/198201/iphone_4s_re... PC World article, is a display that is approximately 300 ppi or better from 12" away.

    Arguably this makes the display on the Galaxy Nexus a "Retina display", or just shy of it. I love my Nexus S display, so this is going be a nice improvement.
    Reply
  • Chinpokomon - Friday, October 21, 2011 - link

    Interestingly, I added JesusR's LG LU6200 to the mix, and found even more reason to disbelieve the SPPI calculations as valuable.

    For the LG LU6200, I got 341.54 calculated PPI, 341.54 Actual PPI, and 583.90 Actual SPPI. My Actual SPPI calculations match the ones from the Subpixel Density Comparison chart, so I believe I've calculated SPPI the same way. That said, as calculated PPI for the iPhone 4S is 325.93 and 341.536 for the LU6200, how is the SPPI 606.39 for the 4S and 583.90 for the U6200?

    As others suggested earlier, I think aspect ratio is a factor here, but when calculating the Actual PPI as I did above, the results give a better representation.
    Reply
  • shimman - Friday, October 21, 2011 - link

    Brian Klug & Jason Inofuentes appear to be biased against Samsung (or not so smart)

    First, subpixel density comparison chart
    what's up with RGBW LCD being better than RGB based screens?

    Second, most of people who did hands on Galaxy Nexus were impressed with the quality of the screen.

    Pentile can create sharper image at same sub-pixel counts than the RGB arrangement

    sub-pixel counts: Galaxy Nexus = iphone4S.
    this means 4.6" Galaxy Nexus > 4.6" with iphone4S resolution.

    If you don't like pentile OLED Super AMOLED HD screen, alternative is LCD screens which will ave grey black, ghosting, low contrast, narrower viewing angle, and less vivid color.

    To me Super AMOLED HD is the best screen for the mobile phone for now.
    Reply
  • Jussi7 - Saturday, October 22, 2011 - link

    I do not see a bias against Samsung. I do see a big gaping hole in your logic.

    If you don't like Pentile OLED, you might try non-pentile OLED first. OLED and Pentile are not a match made in heaven, as shown in Samsung's Super AMOLED Plus screens which were mentioned in the article.

    For example, I have seen Galaxy S (Pentile) and Galaxy S2 (RGB) phones in person. The first screen is way worse (fuzzy, patterns) than the second. The PPI of the first is higher than PPI of second, ergo comparing only PPI between RGBG and RGB and pixel layouts is not solid.
    Reply
  • schriss - Saturday, October 22, 2011 - link

    I own Galaxy S II and I have preordered Galaxy Nexus for it's higher resolution screen. I have now mixed feelings, like I was cheated with fake HD screen, that might have more pixels but they will be fuzzy. What's the point then? If I see more content on that HD screen but it will be fuzzy I will have to zoom in to see same amount of details I see in Galaxy S II?
    That would be huge disappointment, I hope Samsung would never risk releasing a disappointing display and I hope they did use Galaxy S II display as main comparison for the new display and they wouldn't release something that's not an improvement.
    Maybe they will have HD Super Amoled Plus ready for Galaxy S III?
    Reply
  • Filiprino - Saturday, October 22, 2011 - link

    Maybe they will have HD Super Amoled Plus ready for Galaxy S III?

    You can bet on it.
    Reply
  • barnie - Saturday, October 22, 2011 - link

    The reason is simple: Pythagoras' theorem does not work if you are using different measures of distance in the two axes. You're pretending to be measuring a diagonal, but the triangle you are forming to do that has none of the characteristics of the one that actually contains it.

    The simplest realistic approximation involves calculating the area of subpixels and taking a square root to give you a linear measure. It boils down to a sqrt(2/3) factor, or PenTile being <19% worse than RGB in linear subpixels per inch.
    Reply
  • warmongerd - Saturday, October 22, 2011 - link

    Wasnt Brian Klug the guy posting the really biased benchmarks between iPhone 4s and GsII not so long ago? Reply
  • SanX - Sunday, October 23, 2011 - link

    “...people say they never notice the PenTile pixel structure but it is just like a stain in a carpet; once you see it, it is hard to disregard...”

    which means the third figure is needed which shows worst case scenario - for example solid colors DPI for red and blue. Most affected are red, orange, yellow and magenta. Blue is by some reason not.

    Nexus PPI for the red and blue subpixels - 158 which completely explains the "dirty carpet" effect

    here is an example
    http://images.anandtech.com/doci/4761/Bionic-4520....
    Reply
  • SanX - Sunday, October 23, 2011 - link

    That means pentile must be in 600 ppi category for you not to notice "dirty carpet effect" for red part of spectrum colors. That will make ppi for red 300.

    And you will still notice it again when put the cell closer to the eyes, at around at 8" (compared to 12") because perceived ppi will drop to 200

    Not practically doable. This is why Pentile must die.
    Reply
  • SanX - Tuesday, October 25, 2011 - link

    Pentile must be RRGGB. Period Reply
  • RobertStar20 - Monday, October 24, 2011 - link

    The words of this article are nicely balanced, unfortunately the graph is completely meaningless. All of these attempts at analysis completely ignore the fact that the eye sees different colours with different levels of acuity, a concept that is the underlying *basis* of PenTile.

    I've posted some simulated pictures at my site below, which sort-of turn your normal monitor into a PenTile display for the sake of trying to "preview" the Galaxy Nexus screen.

    http://rs20.mine.nu/w/2011/10/the-galaxy-nexus-pen...

    The conclusion? "We'll have to wait and see", exactly as written in this article. But my first impressions are that this is going to be an outstanding screen, in terms of resolution and smoothness.
    Reply
  • SanX - Tuesday, October 25, 2011 - link

    Pentile must be 300ppi RRGGB, 500ppi all other types or just die Reply
  • schriss - Monday, October 24, 2011 - link

    I've read this:
    http://www.phonearena.com/news/Samsung-Galaxy-Note...

    "If you look at the zoomed in 100% crops of pics made from the same distance, the letter borders have deeper jaggies on the Samsung Galaxy S II, as opposed to the ones from the Galaxy Note display, which are closer to the smoothness of the iPhone 4's screen."

    "Cold colors making white appear blueish ever since the Super AMOLED on the first Galaxy S - yes, that exists even in the new HD Super AMOLED , but details were the clearest we've seen on any AMOLED to date."

    And all that was about Note's screen which is bigger than Nexus's. So I think case closed, at least for me.
    Reply
  • SanX - Friday, October 28, 2011 - link

    The guy on this site has no clue what he is doing comparing wrong things.

    The small fonts equivalent to your desktop angular dimensions when you keep the phone at the optimal distances for viewing (25cm and as close as 20 cm, taking in mind also SMPTE or THX requirements which become applicable for the large screens)
    must be compared, not overall picture "clarity" or large areas of screen with large fonts !!! He implies that we will continue zooming our webpages forever while actually with if we will have 400+ppi 5" screens we do not have to because the fonts approach angular dimensions of desktop monitors.

    Think !!!!!!!!!!!!!!!!!!!!!!
    Reply
  • solarein - Wednesday, October 26, 2011 - link

    I also don't understand how you arrived at your numbers. For the iPhone 4, with 3 horizontal subpixels per pixel, you get a horizontal subpixel density of 989.2 and the vertical subpixel density is the same as the pixel density at 329.7. So far so good, but then what did you do with these two numbers to arrive at a subpixel density of 613.3?

    I did a different calculation, I calculated the total number of subpixels of the iPhone (960x640x3) then divided that by the area of the screen (2.91x1.94) to get the number of subpixels per square inch, then took the square root to get the number of subpixels per inch, and got 571. How did you get your number?
    Reply
  • solarein - Wednesday, October 26, 2011 - link

    Ok, I read some of the other comments and saw how you got your number. You took the number of vertical pixels, 960, and the number of horizontal pixels times 3, 640*3, and did

    sqrt(960^2 + (640*3)^2)

    to get...something. And then divided this by the diagonal size of the screen to get your subpixels per inch. The problem is this is nonsense, it's a misapplication of the Pythagorean theorem. This method of calculating "diagonal resolution" and then divide by diagonal screen size works fine when we are talking about square pixels. But it makes no sense when we are talking about non-square subpixels with different horizontal and vertical resolutions.

    To see why this calculation is nonsense, consider this: suppose instead of having 3 horizontal subpixels per pixel, you have 3 vertical subpixels per pixel. In both cases the total number of subpixels is 640*960*3, so the subpixels per inch in the two cases should be the same, right? But using your calculation, for the horizontal subpixels you get
    sqrt((640*3)^2 + 960^2)/3.5 = 613.32

    But for the vertical subpixels case you get
    sqrt(640^2 + (960*3)^2)/3.5 = 842.93

    You get a different number for two screens with the same size and same number of subpixels!

    The right way to do this is how I did it, by calculating the number of subpixels per square inch and then taking the square root. Another way to think about this is simply scaling the pixel per inch by the square root of the number of subpixels per pixel. So for 3 subpixels per pixel of the iPhone, you do 329*sqrt(3) = 569. And for the 2 subpixels per pixel of the Galaxy Nexus you do 315*sqrt(2) = 465. This makes intuitive sense, if you imagine a theoretical screen with 4 subpixels per pixel arranged in a 2x2 grid, you just take the square root of 4, which is just 2, and multiply that by the pixel per inch, and get that the subpixel per inch is exactly double the pixel per inch, which is what you want. Your method both overcounts the subpixel per inch of the RGB grid and undercount the subpixel per inch of the RGBG grid.
    Reply
  • solarein - Wednesday, October 26, 2011 - link

    To elaborate a bit further on why our method is nonsense, it's because you are thinking of the pixel and screen size measurements as scalars, with a magnitude and no direction, when in reality they are vectors with both magnitude and direction.

    When you take the "diagonal resolution" and divide by the diagonal screen size, you are effectively calculating the ratio between two vectors, and recording this as the pixel per inch. It's ok to ignore direction when we are dealing with square pixels, since the direction of the diagonal resolution vector and the diagonal screen size vector are exactly the same. But it's not the case when we are talking about subpixels.

    Consider when you calculated the "diagonal resolution" by taking sqrt((640*3)^2 + 960^2), you got the magnitude of a vector, but this vector is now scaled three times in the horizontal axis. It is no longer in the same direction as the diagonal screen size vector. So when you divide this number by 3.5, the result is nonsensical.
    Reply
  • solarein - Wednesday, October 26, 2011 - link

    This is a ridiculously convoluted way of doing a very simple calculation. What you are doing is just calculating the number of RGB-equivalent pixels of the Galaxy Nexus, then diving that by the area and taking the square root to get the RGB-equivalent pixel per inch of the phone. What you could have done is simply calculate the dimensions of the Galaxy Nexus screen using the diagonal and the aspect ratio, and get 4.05in X 2.28in, and then do:
    sqrt(1280*720*(2/3) / (4.05*2.28)) = 257.9 RGB-equivalent pixels per inch, which you got.

    What you also don't realize is that this number prove the numbers in your second graph to be incorrect. You can get from the number of RGB-equivalent pixels per inch to the number of subpixels per inch by multiplying by sqrt(3) (because the square of sqrt(3) is 3, which is the ratio of subpixels to pixels per square inch of screen). For the Galaxy Nexus you get

    257.9*sqrt(3) = 446.7

    And for the iPhone 4 you get

    329.7*sqrt(3) = 571.1

    These are the correct numbers for the subpixel per inch of each phone. Notice that in this case you use sqrt(3) for the Pentile screen, not sqrt(2) because the 257.9 is the RGB-equivalent number of pixels per inch. If you use the actual number of pixels per inch, 315, then you multiply by sqrt(2) instead to get the subpixels per inch, and you still end up with 446.7.
    Reply
  • solarein - Wednesday, October 26, 2011 - link

    An even simpler way of thinking about this is by noticing that, since the Pentile screen has 2/3 the number of subpixels as the RGB screen, you can pretend that in each dimension it has sqrt(2/3) the number of pixels, so that when you take the area you end up with 2/3 the number of pixels, which is what you want. Once you have this, converting from the pentile PPI to the RGB equivalent PPI is as simple as multiplying by sqrt(2/3). Reply

Log in

Don't have an account? Sign up now