The Galaxy Nexus PenTile display: comparing apples and oranges

There has been a lot of discussion on the ‘net about the infamous PenTile display. Lots of sites are throwing numbers around, comparing the ‘subpixel resolution’ of competing phones such as the Galaxy Nexus and iPhone 4/4S. Some commentary has been insane, comparing apples with oranges and declaring PenTile an abomination regardless of resolution. Others, like this one, are a lot more reasoned, but still seem to ignore the well-proven fact that the eye’s visual acuity depends on the colour in question. Let’s rattle off a couple of undeniably true facts:

  1. Hypothetical 1000ppi RGB and PenTile displays can both look absolutely sharp and perfect.
  2. The acuity of the human eye varies significantly between colours. Acuity is particularly poor with blue light.

Take the following example: these four squares consist of fine lines. From left-to-right, they are: #1: white/black, #2: red/cyan, #3: green/magenta, and #4: blue/yellow.

Stand well back from your monitor, until you can’t tell the lines apart. Now slowly move back in. Notice how the blue/yellow (#4) lines look almost the same as white/black (#1), and red/cyan (#2) look very similar to green/magenta (#3)? Very roughly speaking, this is because our eyes smear out blue light slightly, so we can’t tell the difference between white lines (yellow+blue), or alternating yellow and blue lines. Similarly, the only difference between red/cyan and green/magenta is where the blue lies.

I started this not-terribly-scientific experiment with the “hypothesis” that the eye had a higher acuity to green than red and blue. This is apparently the reason that is used to justify PenTile’s use of twice as many green subpixels as red or blue. However, although my pictures above suggest that green is more visible than blue, my eyes are more-or-less equally acute with respect to red and green light. This is backed up by at least one actual proper study, as well. [Interestingly, the original PenTile technology was RRGGB, which makes much more sense!]

So, we can conclude that the resolution of red subpixels might be the limiting factor for an RGGB “Retina” PenTile display, whereas the resolution of green subpixels might be the limiting factor for an RGB Retina iPhone 4 display. Comparing the two of these numerically is very difficult and highly subjective. The sharpness with which the eye can see these colours is different, and even this difference varies according to the brightness of the image, and the individual. To see numbers calculated with no apparent regard to these factors and presented to five significant figures is just laughable.

The closest you can come to a meaningful comparison is stating what sort of displays would be equivalent if green sub-pixel resolution was all that mattered, and what would be equivalent if red/blue sub-pixel resolution was all that mattered. The linear green sub-pixel resolution is the same for both RGB and PenTile displays, because both feature full green sub-pixels. The linear red/blue sub-pixel resolution of PenTile is compromised by a factor of sqrt(2), because red pixels are diagonally separated. So, here we go:

The effective equivalent RGB resolution of a 1280×720 PenTile display probably lies somewhere between 905×509 (if red/blue is the limiting factor by far) and 1280×720 (if green is limiting factor by far.) That is, somewhere between the-highest-resolution-ever-seen-on-a-smartphone (except the iPhone 4S by a hair) and the-highest-resolution-ever-seen-on-a-smartphone (by a mile, although some [dirty LCD] RGB 720p devices have been launched since.)

And we’ve come all this way without mentioning display size. The bigger the display, the bigger the font can be and the further you can comfortably hold the device from your eyes. A bigger display should therefore mean a lower linear resolution limit before the display be reasonably considered a ‘Retina’ display.

The proof of the pudding is in the eating

My conclusion? It is virtually impossible to pre-judge a display that we haven’t even seen before. I would love to hear subjective comments from people who have seen the device in person; otherwise, I’m not interested.

But how should I pre-judge this display?

Well, OK, if you must know. Having seen a complete dearth of any meaningful analysis on the web, I decided to sit down and try and figure out how good this display is. Any calculations were out of the question because, like everyone else taking part in this debate, I’m simply not qualified to comment on what the eye can and cannot perceive. At least I’m brave enough to admit it.

So the next best thing is an attempt at simulating a 720p PenTile display. I decided a normal, RGB computer monitor is a pretty good starting point. PenTile is characterised by one of R or B being missing from each pixel. So, for each pixel, I either set the blue component to zero, and doubled the red component, or vice-versa. The end result is shown in the last panel below.

I concede that the second and third panels look markedly different, but they do share the same fundamental properties:

  • Green subpixels are everywhere — one for each pixel.
  • Red subpixels are in a checkerboard/”diagonal” pattern.
  • And blue subpixels are the same as red subpixels, except in complementary positions.

So, fundamentally, they both enjoy full linear resolution in the green channel, and suffer from a linear resolution compromised by sqrt(2) in the red and blue channels.

[ For the techies out there, I mostly took care of the sRGB gamma curve (well, as close as I could since sRGB doesn’t follow a perfect gamma curve.) For the non-techies out there who are wondering what I’m on about, if you want to make a pixel output exactly half (or double) the amount of light, you cannot just halve (or double) the luminance value — not even close. I wish it were that simple, but look at sRGB to find out the horrifying truth. Anyway, I digress. ]

The sample images! Finally!

The images below must be viewed at 100%; they’re completely meaningless if expanded or reduced.

One catch is that in order to double the brightnesses of the red and blue subpixels, I had to start with a dim image that was below 50% brightness everywhere. This is why the images below are quite dim. The left side of the image is my PenTile simulation, the right side is a completely standard (dimmed) picture. Now, because you’re monitor isn’t 316 ppi, you need to sit well back from the screen to get an accurate idea of how the display looks.

A ruler helps — the Galaxy Nexus display is, by my calculations, 10.3cm wide. So, hold a ruler at arm’s length, and walk back from the screen until the image “looks 10.3cm wide.” Now you know what the display looks like at arm’s length. If you want to hold the “phone” closer, bring the ruler closer, and move towards the screen until the left of the image is at 0cm and the right of the image is at 10.3cm on the ruler.

If you’re so close to the monitor that you can’t see the ruler in focus, then you’re cheating!

Ignore any colour cast you see on the image — this is to do with the gamma of your monitor (and my monitor, for that matter) and doesn’t represent any real-world property of PenTile displays (although colour rendition is cited as an issue with PenTile displays, see below.)




My conclusions?

When I stand/sit just a little way back from the monitor, the images above become crisp and beautiful, even while they’re still very big. On my 94 ppi monitor, the images look absolutely perfect and smooth by the time I’m 80 cm from the screen. On the 316 ppi Galaxy Nexus, that corresponds to 23.8cm, which is less than the 12-inch (30.5cm) metric used by Apple to justify the iPhone’s ‘Retina’ name. More to the point, I’d say that’s unnaturally close. And this is all based on my weird simulation, which definitely ought to be more “speckly” than the real thing, considering the empty black sub-pixels everywhere.

Keep in mind that the right-hand side of these images represent a hypothetical 720p full RGB display, which simply doesn’t exist on any AMOLED smartphone, nor any iPhone*. Whether the Nexus display will be like the iPhone display in terms of being pretty close to indistinguishable from perfection at any face-to-phone distance is still hard to tell. So let’s just all wait for the reviews from people who have seen this phone in person, shall we?

* Some true RGB 720p devices have been announced, such as the HTC Rezound. However, they use LCD technology, presumably with the accompanying poor contrast ratio and performance in bright conditions.

Have we talked about colour rendition yet?

Problems such as poor colour rendition are not necessarily endemic to PenTile displays. It takes time for any new technology to be perfected. RGB has been around since 1953, PenTile RGBG since 2009. Every new display is a brand-new product; you can’t put a display in a photocopier at 50% to get a higher resolution, you need to re-design every aspect of the display to handle the new dimensions. Thus, every different resolution, every different screen size, is a whole new part; a new opportunity to fix problems associated with previous generations of the technology.

It stands to reason that a PenTile display’s colour balance might vary due to viewing angle because the sub-pixels vary in size and shape, which could clearly lead to different emission patterns for each colour of light. Is that a tricky problem to fix? Yes. Should we just blindly assume that the problem hasn’t been fixed? Of course not. In that context, check out the follow short-sighted update to a post:

Update: We just added AnandTech’s analysis at the More Coverage link. They point out that if you like the pixel density on the GS II, you should be happy with that on the Galaxy Nexus — although they don’t address color rendition.

I think this little comment sums up the absurdity of this whole debate nicely — linking to a site which offers some relatively meaningless numbers which ignore screen size, expected viewing distance, and varying subpixel resolutions of different colours, then conceding that the subpixel resolution is probably actually very good, only to complain about color rendition — which is a complete unknown at this point. True, it’s not actually a complaint per se, but reading the entire post leaves a very sour taste in your mouth, when the message should be “just hold on a second, there’s one or two things we need to check before we declare this the best display ever.” Which to me, is just a statement of the obvious.

Share
Leave a comment

15 Comments

  1. Alex

     /  October 25, 2011

    Very nice and well written analysis. That you for the simulated images, that saved me from writing my own program to do so. I had another idea of how to simulate pentile on a regular LCD, not sure if it is entirely correct: one could simply take the mean value of each two consecutive pixels form the blue channel and place that in both pixels; the same should be done, with a one-pixel shift, for the red channel as well, and repeated shifted by one pixel on each consecutive line.

    Reply
  2. chee

     /  October 25, 2011

    wow, this is really well written. I actually understand what you are saying.

    Reply
  3. Robert

     /  October 25, 2011

    Thank you for the kind comments!

    Alex, your suggestion for an alternative way of simulating PenTile is very good — it shares the same fundamental properties of having full resolution in the green, and resolution in red and blue both being compromised by sqrt(2). However, there is a bit of a problem — your suggestion has all of the red sub-pixels glowing, which will make the distribution of red light much smoother (less “speckly”) than a real PenTile display. The same goes for blue.

    Since the purported disadvantages (graininess/specklyness) of PenTile are mostly due (in my opinion) to red being completely absent from every second pixel, I think it’s important to try to reproduce this disadvantage of PenTile.

    It’s true that my simulation actually unfairly goes slightly too far in adding the speckly look to the picture — but I think this is the “right way to be wrong”, given that I still conclude that “720p PenTile sounds pretty sweet”.

    Reply
  4. Joseph

     /  October 27, 2011

    Loved the article, really put out some good info on the topic. What would the advantage of using a PenTile matrix then if there is all this out cry from people, why not just use the standard RGB matrix display? Also, which do you think is better in your opinion..is there an advantage to one over the other?

    thanks

    Reply
    • Robert

       /  October 27, 2011

      Hi Joseph, thanks for the kind comments!

      A PenTile matrix can, in theory, give a better quality image for a given number of sub-pixels. Given that the manufacturer pays for each sub-pixel (since each sub-pixel is an actual, physical LED or LCD segment), and the consumer wants a good quality image, this is a win-win for all. A full RGB display wastes your money on a relative over-abundance of blue sub-pixels. This is the theory, at least (the bit about the blue sub-pixels is certainly true.)

      The only reason for the out-cry is the fact that a PenTile display of a given resolution and ppi is not as good (strictly clarity/smoothness-wise) as an RGB display of the same resolution and ppi. However, many people seem to be incapable of realising that this comparison only holds true when the resolutions and ppi are the same. Once the resolution/ppi of a PenTile display is more that 41% higher than an RGB display (as in the case of Galaxy Nexus vs Galaxy S II), the PenTile display definitely ought to be better (strictly clarity/smoothness-wise.) If the PenTile display is between 0 and 41% higher resolution than an RGB display, the question of comparing them (clarity/smoothness-wise) becomes very difficult (because it depends on the properties of the human eye) and subjective, and I’m not qualified to comment, but neither is anyone from Anandtech or Engadget by the looks of things.

      In short — do not judge a display on whether it is PenTile or RGB. All of the purported disadvantages of PenTile are misplaced criticisms of specific past PenTile displays, and are not intrinsic to the PenTile sub-pixel layout at all. As I said in the article, every new display, PenTile or not, is a brand new device, and assuming it is identical to past displays is just foolish. So wait for proper, detailed reviews to be published, and compare screen A with screen B, not subpixel layout A with subpixel layout B. [ I think few people will disagree with me when I say that AMOLED is much better than LCD. Better power efficiency, colour vibrance, and contrast ratio. ]

      If you’re thinking about the Galaxy Nexus, I believe it is going to be a stunning, colourful, sharp, smooth display. Smoother and sharper than any other Android phone. If you are more interested in colour accuracy (e.g. for photographic work) than sharpness and smoothness, then it’d be prudent to await further verdicts from reviewers who a) know what they’re talking about and b) have held the phone in their hands. Because the sub-pixels of the Galaxy S II are uniform in shape and size, they might emit light in a more uniform pattern, so colour might be more consistent with different viewing angles on the S II. But maybe the 720p PenTile display is different, and has all the subpixels the same size and shape, in which case the Nexus display will be unquestionably the best display ever. But here I am, drifting into the sort of crazy speculation that this post is all about avoiding. Just wait for proper reviews :-).

      Reply
      • Brandon

         /  December 6, 2011

        Hey,
        Great article. You’ve put some genuine work into explaining the issue so that everyone can understand.

        I notice your reply here has this statement
        ” If you are more interested in colour accuracy (e.g. for photographic work) than sharpness and smoothness, then itโ€™d be prudent to await further verdicts from reviewers who a) know what theyโ€™re talking about and b) have held the phone in their hands.”

        I think if somebody is very worried about color accuracy for photo work they might want to consider a device that isn’t a smartphone. Wouldn’t you agree? I always here this argument about AMOLED displays being oversaturated(always claimed by users of TN panel monitors I might add) and I have to wonder what kind of color sensitive work they hope to do with their smartphone.

        Reply
        • Robert

           /  December 6, 2011

          Hi Brandon, thanks for the kind comments! You raise a good point, smartphones are not going to be used for colour-critical work. My comment was more referring to the question of whether or not colours changed according to viewing angle, which was still a bit of an unknown.

          I would break the statement up though:

          a) OLED displays would be absolutely brilliant for a PC monitor. The fact is that the primaries of an OLED display are (generally?) very vivid — but the display can still make grey and everything in between. As long as the software knows the characteristics of its display, vivid primaries (read: a wide gamut) only expand the palette (gamut) of colours available to the display. This is where colour calibration devices and colour-space-aware applications and operating systems come in. Thus, I would love an OLED display for my photography work — even the fact that OLED displays can slightly shift their white point over their lifetime can be perfectly compensated for with calibration.

          b) But yeah, doing all of the above on the smartphone seems a little crazy. ๐Ÿ™‚

          Reply
  5. Eduardo

     /  November 5, 2011

    Actually a 720p full RGB display does exist for in a smartphone. There’s an LG device in South Korea that has one, and I believe an HTC device launching in the next few days in the US.

    Otherwise an excellent attempt to discuss the issue rationally – though unfortunately we’re missing the key test of black text on a white background. I’m assuming you didn’t do this because you couldn’t determine a suitable sub-pixel anti-aliasing for your virtual-PenTile scheme?

    Reply
    • Robert

       /  November 5, 2011

      Thanks for the feedback on the those other phones; I’ve just patched up the article accordingly. For the record, I don’t necessarily consider those phones too much of a threat because they’re LCDs; as much as I love high resolution, I have to give priority to contrast and readability. Of course, it’s hypocritical of me to presume that the new LCD screens will perform below par in these respects, so my ultimate conclusion of “wait for the reviews” still stands.

      Doing a simulation of text didn’t really cross my mind, but there are a few reasons why I’m not keen to do so:

      1. It would be very difficult! Like, more than ten minutes! For me to do it properly, we’re talking about me re-implementing Microsoft’s ClearType from scratch.
      2. Furthermore, even if I did figure it out for my screen, it wouldn’t work on BGR monitors, or monitors rotated 90 degrees. These images would suffer from bad colour fringing. These problems are actually present on my simulated photos, but they’re so subtle that you really have to search for them.
      3. We already have word (from people who have actually interacted with this device) that the text on this device is amazingly crisp and clear, no mention of the tradition PenTile complaints.
      4. This is a much weaker point, but given that my simulation has significant black voids within it, and the picture still appears perfectly smooth from a sensible distance, it seems unlikely that the display would somehow still fail to present indistinguishable-from-perfect text.

      In other words, I feel like I’ve more-or-less made my point already. ๐Ÿ™‚

      Having said all this, I might get around to trying it if I get any huge chunks of spare time!

      An interesting corollary from all this is that a PenTile display must be paired with an OS that understands PenTile and adjusts its “ClearType” antialiasing accordingly, or else the display will suffer from exactly the sort of colour fringing that has been characteristic of PenTile displays in the past. With all the sentiment from Google and Samsung that their engineers worked in the same building during the development of this phone, perhaps Android ICS has finally slipped this feature in under the radar? More juicy speculation.

      Reply
  6. Robert,

    Excellent. Its is refreshing to see our work being analyzed in a more systematic fashion.

    Full Disclosure, I am the CEO of Nouvoyance, the firm that developed and supports PenTile technology.

    A few comments: We do simulate layouts and algorithms using LCD panels, using multiple pixels to represent subpixels, then view them from a distance. There are also models of human vision which we use to evaluate them as well.

    Yes, my original PenTile layout used equal amounts of R & G subpixels. It is also more information efficient, but requires higher resolution than PenTile RGBG to effectively match the human retinal algorithms. It is also more difficult to manufacture, not using rectilinear subpixels and grid points .

    One other thing to keep in mind about the RGBG layout… the loss of resolution in pure red occurs on the diagonal, not the horizontal and vertical. The human vision system similarly has lower resolution in the diagonal, compared to the horizontal and vertical… so trying to compare resolutions based on simple X by Y matrix “equivalent” is unwarranted and misleading. Another think to keep in mind is that for images that remain clear of the saturated red point have significant amounts of green, thus they allow for full image reconstruction of the luminance signal, while the reduced number of R and B reconstruction points is acceptable because they are mostly there for chroma reconstruction, for which the human vision system has a MUCH lower resolution.

    I’m also heartened to see you reach the insight that when we use PenTile RGBG for OLEDs, we do so to obtain higher resolutions than would currently be reachable with the RGB Stripe architecture. Simply stated, PenTile allows better images, for the number of subpixels actually present on a given panel, than RGB Stripe, as you pointed out. This has always been our goal, our value proposition, in developing the technology.

    Reply
  7. Sharky

     /  May 11, 2012

    Nice article! What about the 720p display of the Sony Xperia S? I saw it in action and was very impressed with the contrast.

    Reply
    • Robert

       /  May 19, 2012

      I haven’t seen that display in person, but it’s an LCD display, which normally means not-so-black blacks. But perhaps Sony have done some magic to fix this.

      The first (and only) review that I’ve just had a look at suggest that the viewing angles of the Xperia S are “sub-par”. Furthermore, although the contrast ratio of the Xperia S is rated a “very good” 1038:1, this compares to ∞:1 for any AMOLED screen. Whether this matters is up to the individual.

      Reply
  8. mirak

     /  August 30, 2012

    I totally disagree with this article.
    Pentile creates aliasing on the horizontal and vertical lines, and you can defenitely see that.
    This is so obvious to my eyes.
    I had a HTC Desire and really the only thing that saves Pentile is the contrast and colors of the AMOLED.
    I have a galaxy S2 and it’s really the best display specially because it’s not pentile.

    Reply
    • Robert

       /  August 30, 2012

      The HTC Desire is a 480×800 Pentile screen, the Galaxy S2 is a 480×800 RGB screen. If you read my article carefully, you’ll find that I conclude that at a fixed resolution, the RGB screen will be sharper than a Pentile screen.

      You’ve completely missed the point of the article — I’m pointing out that if a Pentile screen has a higher resolution than the RGB screen, then it’s dumb to just say “oh, it’s Pentile, therefore it’s worse.” That conclusion only holds true if the nominal resolution is the same for both devices.

      Try picking up a device with a higher resolution that is only possible today with Pentile on AMOLED, like the Galaxy Nexus or Galaxy SIII. Those Pentile screens are vastly superior to the RGB SII.

      Reply
  9. Khawar Nadeem

     /  November 6, 2012

    I love you. This was hands-down the best analysis of the screen I’ve read. Thank you for teaching me a lot of new things today ๐Ÿ™‚

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *