Apple Watch Pixels | Hacker News Reader

Apple Watch Pixels

prometheus.med.utah.edu

118 pointsdavidbarker36 comments

Sub-pixel rendering on a pattern like that must be complicated?

Sub-pixel rendering isn't really worth it on high DPI displays like this. It also doesn't work that well with iOS's CoreAnimation compositing model. Blending sub-pixels properly across transparent layers in particular is difficult.

Does the apple watch downscale like their phones? Can apple implement subpixel downscaling? It seems like an easy win if already downscaling abbreviations if you can do it in hardware.

Hardware gets a flat image. It can't differentiate between foreground text and background. It has to be done by the compositing layer.

I'm not convinced that is necessary. Here's my first thought. If I'm given a higher resolution image and am to downscale, knowing it will be displayed on a particular screen, then I would sample the blue pixel not from the "center of the pixel" but from the position of the blue dot in the screen. Similarly for red and green pixels. Couldn't an approach like that give subpixel precision?

If you're guaranteed that the image is displayed properly aligned to pixel boundaries, then doing that might do something, but it's probably not worth it.

Which it probably will be, but it's certainly not guaranteed.

The Apple Watch has 290/302 dpi which barely qualifies as high DPI. 290 is just below their own stated Retina threshold. Anyway, I hope they'll have a better screen for 2nd or 3rd gen watch.

Edit: The numbers are wrong, it's 326 dpi for both because I couldn't be bothered to check the math of the retina display wikipedia article.

Both screens have equal pixel densities and have 326 ppi (the larger screen has consequently more pixels). Where do you get your 290/302 ppi from? That’s just wrong.

The display of the 38 mm watch (38 mm refers to the height of the whole watch body, including bezel – it’s ill-fitted for calculating pixel density) is 21.22 mm wide (0.8354 inches) and has 272 horizontal pixels. That’s 325.6 ppi.

The display of the 42 mm watch is 24.34 mm wide (0.9583 inches) and has 312 horizontal pixels. That’s also 325.6 ppi.

That’s the same pixel density as the iPhone (excluding the 6+ with a higher pixel density and older non-retina iPhones).

Now you can argue about whether typical viewing distances for watches are closer or farther than for phones (whether or not a screen is “Retina” is always a function of the typical viewing distance and the pixel density, never pixel density alone and there is no single device independent pixel density that is “Retina”), though I actually think on balance they are probably slightly farther or the same.

You're correct, someone screwed up in Wikipedia (the Retina display page) and I didn't bother to check the math.

> Sub-pixel rendering isn't really worth it on high DPI displays like this.

[Citation Needed]

The whole point of sub pixel rendering is to make text look like it's higher resolution. Once you get to a sufficiently high DPI there's no benefit to it.

Yes. However, the question is "are apple displays "sufficiently high DPI"". They are higher DPI than most other displays - but that is not the same thing.

My answer is "yes, they are sufficiently high DPI."

a) if it truly is "retina," then your eye can't distinguish pixels smaller than it, so subpixel text is meaningless.

b) Laser printers started at 300 DPI, at 1 bit of color and even now have pretty much settled on 600 DPI, still usually 1 bit of color. Retina displays are about 300 DPI with 8 bits of color (in black and white). So, roughly comparable to a laser printer.

You're forgetting movement. Which is not an issue with printers, but is an issue with displays. Such as when scrolling and such. You can detect the difference between moving at 5 inches/sec and, say, 7.5 inches/sec quite easily.

With no subpixel rendering, assuming a framerate of 60fps, you can move only at the following speeds: 5 inches per second (1 pixel / frame), 10 inches per second (2 pixels per frame), 15 inches per second (3 pixels per frame)...

If you're running at 60 FPS and don't mind judder above 30FPS (hideous, but some people apparently don't mind it. I am not one of those people. On a related note: for some reason PM occasionally tears webpages when scrolling, and it drives me crazy), things are a little better, but not much. 2.5 inches / second (1 pixel / 2 frames), 3.75 inches / second (1 pixel then 2 pixels, alternating), 5 inches / second (1 pixel / frame), 6.25 inches / second (2 pixels then 3 pixels, alternating)...

You cannot maintain a scroll speed slower than 5 inches per second (approximately: 300dpi @ 60fps. If you compromise for 30fps, that's 2.5 inches per second instead) without dropping your effective framerate without subpixel rendering. This is an issue all over the place.

As such, yes, subpixel rendering is still required even with retina displays.

Yes, you need the ability to draw content at fractional pixel offsets to get smooth motion. That's different than the "subpixel rendering" the original post was talking about, though, which is a way to draw sharper shapes (usually text) by addressing the red, green and blue "subpixels" individually: https://en.wikipedia.org/wiki/Subpixel_rendering

Except that they are functionally identical. In order to draw at subpixel offsets you need subpixel rendering (or rather: if you have drawing text at subpixel offsets without proper subpixel rendering you get really bad artifacts)

Your claim that you get really bad artifacts on a retina display when you draw lines properly is also needs citation. I don't know if the retina display has enough pixels to meet the claims either, but it seems you're confusing anti-aliasing with subpixel rendering.

https://en.wikipedia.org/wiki/Spatial_anti-aliasing#Examples

https://en.wikipedia.org/wiki/Subpixel_rendering

It is possible to show, however, that if a scene has a band-limited detail property and the join response of your visual system, your eyes' point spread function and the pixel geometry resembles a Nyquist pulse, then indeed there is a density (which would be equal to the density where 1 pixel is covered by 1 sampling element in the retina) which is sufficient for perfect reconstruction of the image, even if it is moving.

> It is possible to show, however,

Yes. However, the retina display is not there yet.

All you need is regular antialiasing, not subpixel antialiasing. To be precise about what I mean: drawing a black line on a white background with regular antialiasing will produce only black, white and gray pixels. Subpixel antialiasing will produce colored pixels. What I'm saying is that you don't need the colored pixels to produce smooth motion.

Apple displays aren't particularly denser than those of other manufacturers. Which stands to reason, as they don't actually make displays. They purchase them from the same market other vendors do.

When the (whole) pixels are already small enough to the extent of being almost invisible to the naked eye, there's no point in trying to squeeze more resolution out by using individual subpixels.

You're forgetting movement. See my other comment: https://news.ycombinator.com/item?id=9849223

The whole point of Retina displays is that the pixels are below the level of the average eye's ability to discern them individually. That's where the Retina monicker comes from. Asking for a citation for a well known issue, that's been widely reported for years, isn't going to win you many friends.

"I am infinitely more comfortable with an Apple device collecting personal information than I am with some of the competing products from other companies"

I couldn't help but notice that quote, either. I wonder why the author found it necessary to put that in on an article about the pixels of the Apple Watch? I don't see that information doing much besides inciting a flame war.

It's a footnote. I guess the author is going for a "stream of consciousness" around Apple watch pixels. (Don't get me wrong, I don't think this is a positive contribution to the article)

So it's a PenTile display [0].

[0] https://en.wikipedia.org/wiki/PenTile_matrix_family

From what I've seen, PenTile appears to be mainly a cost-reduction trick that works well for displaying photographs and other images with gradual variations, but looks horrible for text and sharp, straight lines. A similar pattern has been used in cheap cameras for many years:

http://www.dansdata.com/images/cdv350/screenb560.jpg

The Apple Watch pixels are not in the traditional stripe pattern, but there's still 3 distinct subpixels for each pixel and they're arranged in a square pattern. The unequal subpixel sizes are due to OLED brightness fade; the blues wear out the fastest, while the reds last the longest:

https://en.wikipedia.org/wiki/OLED#Disadvantages

The unequal subpixel sizes are due to OLED brightness fade; the blues wear out the fastest, while the reds last the longest:

Does it vary the intensity as it ages? Otherwise, it seems like the displays would start out way too blue before eventually becoming "accurate". (And then going on to become too red, of course.)

Edit: Oh, or maybe having a bigger blue subpixel means they can run the blues with less intensity, thus extending their life?

Don't the latest Samsung phones (S6) still use pentile? I don't think those look bad at all.

They look fine now because the resolution of the screen is so high. Previously you could definitely notice issues with it.

Not really, it's still RGB, because you still have the same number of red, green and blue cells. PenTile has twice as much green pixels than red or blue (RGBG).

The blue pixels are twice as large as the green and red though. I've never seen anything like it, personally

It looks like they made the miniaturize the pixels to increase contrast and make more room for pressure detectors. However, since blue pixels can't be made as bright as the other colors they have to be bigger than the other pixels.

Has anyone actually confirmed that the display has physical strain/pressure sensors, or is Apple just using the diameter of the capacitive touch point as a faux-pressure calculation?

Yeah, I find the pressure sensitive API in Android is good enough, it just uses the diameter of the touch. It's good enough for varying my pen strokes in a sketching app.

I haven't tried a Note with a true pressure sensitive pen though.

Reply on news.ycombinator.com