Optics help
Oct. 21st, 2010 03:42 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
tahnanAll right, I've hit a limit on my expertise, to the point that I don't even know what terms to Google. But there are enough font geeks, physicists, and lighting designers who read this that someone should know what's going on.
Consider the following diagram, consisting of four instances of the capital letter X in the Georgia font. One of them is oriented correctly (the others being rotations or reflections of the original), and I know that I personally have a pretty instinctive sense of which one.
Of course, it helps that I put it in the upper left, but even so. If you compare the X in the upper left to the X in the lower left, the latter probably just feels wrong. That's an effect of the way serif fonts are drawn: \-slashes are thicker than /-slashes, so if you flip the X, it'll look a little odd.
But what's the difference between the left-hand Xs and the right-hand Xs? In terms of the weight of the strokes, the upper two are identical; indeed, if you superimpose one over the other, they match up...well, not perfectly, but pretty close. When I look at the two, however, the right-hand X looks blurrier to me, like there are little blurred bits of color at the edges; that's true on the top and on the bottom. The reason is that the right-hand Xs weren't created the same way as the left-hand Xs: the right-hand Xs are 180° rotations of the left-hand Xs.
When I noticed this, I wondered why it was. There's a pretty standard wordplay premise that some letters are symmetric, either left-to-right (A, H, T, X...) or top-to-bottom (E, H, X...) or through a 180° rotation (H, S, X...). Of course, how symmetric the letters are depends on how they're written—clearly it assumes a sans-serif font, for instance. But even in a serif font, an X should look the same when put through a 180° rotation; why do there seem to be little blurry colors?
The answer, of course, is that there are little blurry colors. The reason they're there on the rotated Xs is that they're there on the original X as well, which you can see if you enlarge the image. (You can also see it fairly dramatically if you look at two different screenshots taken from MS Paint, in which I've been doing all this manipulation: one while inputting text, and one after inputting text, once the program has inserted the text into the image.) Along the left side of each letter, typed normally, is a one-or-two-pixel-thick line of brown pixels; along the right side, a range of bluish pixels. When the X is rotated, the brown ends up on the right and the blue on the left.
All right, fine, but my question now is this. Why is it that this thin line of color is practically invisible when we look at a letter normally, and yet it stands out so notably when we flip it? Is there some fact about the eye that brown-then-black-then-blue looks black, whereas blue-then-black-then-brown looks blurry?
In the second sentence of this post, I mentioned lighting designers. That's because one of the very few things I remember about stage lighting (other than "Damn those things are heavy") is that, if you want to wash the stage with white light, you don't want to use white light. White light is glaring and washes out the actors. Rather, you want to simulate natural light by using two very pale colors, often a warm on the left and a cool on the right; the ones I learned go by the memorable Roscolux names of Bastard Amber and No-Color Blue. That is, brown on the left and blue on the right. It's like a conspiracy! Except that it's presumably not a conspiracy; it's presumably something about the way the eye parses color that makes it natural to have a warm/brown on the left and a cool/blue on the right.
Unfortunately, I have no idea what it is, nor do I even know how to search for it. All I've really learned is that typing "bastard amber" into Google gets you, to my surprise, no Roger Zelazny fanfiction. Which is neither here nor there. But surely someone in DWland (or LJland) can explain this.
Consider the following diagram, consisting of four instances of the capital letter X in the Georgia font. One of them is oriented correctly (the others being rotations or reflections of the original), and I know that I personally have a pretty instinctive sense of which one.
Of course, it helps that I put it in the upper left, but even so. If you compare the X in the upper left to the X in the lower left, the latter probably just feels wrong. That's an effect of the way serif fonts are drawn: \-slashes are thicker than /-slashes, so if you flip the X, it'll look a little odd.
But what's the difference between the left-hand Xs and the right-hand Xs? In terms of the weight of the strokes, the upper two are identical; indeed, if you superimpose one over the other, they match up...well, not perfectly, but pretty close. When I look at the two, however, the right-hand X looks blurrier to me, like there are little blurred bits of color at the edges; that's true on the top and on the bottom. The reason is that the right-hand Xs weren't created the same way as the left-hand Xs: the right-hand Xs are 180° rotations of the left-hand Xs.
When I noticed this, I wondered why it was. There's a pretty standard wordplay premise that some letters are symmetric, either left-to-right (A, H, T, X...) or top-to-bottom (E, H, X...) or through a 180° rotation (H, S, X...). Of course, how symmetric the letters are depends on how they're written—clearly it assumes a sans-serif font, for instance. But even in a serif font, an X should look the same when put through a 180° rotation; why do there seem to be little blurry colors?
The answer, of course, is that there are little blurry colors. The reason they're there on the rotated Xs is that they're there on the original X as well, which you can see if you enlarge the image. (You can also see it fairly dramatically if you look at two different screenshots taken from MS Paint, in which I've been doing all this manipulation: one while inputting text, and one after inputting text, once the program has inserted the text into the image.) Along the left side of each letter, typed normally, is a one-or-two-pixel-thick line of brown pixels; along the right side, a range of bluish pixels. When the X is rotated, the brown ends up on the right and the blue on the left.
All right, fine, but my question now is this. Why is it that this thin line of color is practically invisible when we look at a letter normally, and yet it stands out so notably when we flip it? Is there some fact about the eye that brown-then-black-then-blue looks black, whereas blue-then-black-then-brown looks blurry?
In the second sentence of this post, I mentioned lighting designers. That's because one of the very few things I remember about stage lighting (other than "Damn those things are heavy") is that, if you want to wash the stage with white light, you don't want to use white light. White light is glaring and washes out the actors. Rather, you want to simulate natural light by using two very pale colors, often a warm on the left and a cool on the right; the ones I learned go by the memorable Roscolux names of Bastard Amber and No-Color Blue. That is, brown on the left and blue on the right. It's like a conspiracy! Except that it's presumably not a conspiracy; it's presumably something about the way the eye parses color that makes it natural to have a warm/brown on the left and a cool/blue on the right.
Unfortunately, I have no idea what it is, nor do I even know how to search for it. All I've really learned is that typing "bastard amber" into Google gets you, to my surprise, no Roger Zelazny fanfiction. Which is neither here nor there. But surely someone in DWland (or LJland) can explain this.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
(no subject)
Date: 2010-10-21 09:53 am (UTC)I've also heard the conventional theatrical wisdom that stage right (to the audience's left) is somehow inherently 'warmer' than stage left, although I have no idea what in human perception or cognition might make this so. But I think your experience with the Xen has a different explanation: subpixel rendering. Basically, modern font-smoothing technology takes advantage of the positioning of the red, green, and blue components within individual pixels to gain more precise control over the outlines of characters. So if this is what's going on, what matters is not that your eyes are seeing brown on the left and blue on the right, but that at the left edge of the black letter, the right-hand sides of the pixels are dark, and at the right edge of the letter, the left-hand sides of the pixels are dark. If your monitor, like mine, has pixels arranged RGB, then the pixels on the left of the X should be showing mostly R (and maybe also G), and the ones on the right should be showing mostly B (and maybe also G)—i.e., reddish-brown on the left and greenish-blue on the right.
stage right warmth
Date: 2010-10-21 10:53 am (UTC)(no subject)
Date: 2010-10-21 11:24 am (UTC)The typographical term that you're looking for, though, is "stress." In many fonts based on the Trajan column there is "uneven stress" in which strokes in different directions have different widths. As an extreme example, consider the common Hebrew prayerbook fonts, in which horizontal strokes are quite wide and vertical strokes are very thin, or in English think of Bodoni.
This is an artifact of the fact that our writing started off with nib pens, which naturally draw a wider line in one direction than in another, and with the usual way a right-handed person drags such a pen across the paper while writing certain letterforms.
(no subject)
Date: 2010-10-21 02:46 pm (UTC)(no subject)
Date: 2010-10-21 04:43 pm (UTC)I'm still curious about whether there is something to different colors implying different directions, as with the theater lighting. It could be just a theater convention that we've all internalized, but it could also have a natural basis. Something to do with the path of the sun across the sky or the very slight diffraction that occurs at the edges of shadows, maybe, I don't know.
(no subject)
Date: 2010-10-21 08:08 pm (UTC)More to the point, though, it's not that I look at it and say "Whoa, upside down"; I look at it and say "Whoa, blurred colors". While the ex-height may have some effect, it's definitely not the thing I'm noticing here.
(no subject)
Date: 2010-10-21 11:31 pm (UTC)(no subject)
Date: 2010-10-21 11:57 pm (UTC)Followup question: what happens when you type white text on a black background?
(no subject)
Date: 2010-10-22 12:24 am (UTC)(no subject)
Date: 2010-10-22 09:09 am (UTC)Couldn't it be that it's doing it right for Tahnan's monitor, but wrong for yours? If your pixels have their colours in a different configuration, then what looks right on one screen will look wrong on the other.
(no subject)
Date: 2010-10-22 12:27 am (UTC)(no subject)
Date: 2010-10-22 12:14 am (UTC)The colors themselves are certainly the result of subpixel rendering (a.k.a. subpixel antialiasing). If you obtained the rotated images by taking a screen shot and then rotating it, it's unsurprising that the rotated version looks wrong, because subpixel rendering is designed to take advantage of the R-G-B ordering of pixels.
(no subject)
Date: 2010-10-21 06:59 pm (UTC)