Why are subpixel values used in the APCA Non-Text Minimum Thickness table?

[angyb]

In the (draft) Semantic Non-Text lookup table found on this issue:
#97

I noticed that many of the recommended widths in the table are subpixel values (e.g., 3.65px for Lc40, or 1.5px for Lc75), and I was hoping @Myndex could clarify something for me:

QUESTION 1: For accessibility, is it acceptable to use fractional pixel widths given that subpixel rendering can cause soft edges or visual blur, especially on lower-density displays?

I would think we’d want to round to whole pixels to maintain sharpness and fidelity.



QUESTION 2: Next, why does this table go as low as Lc5 and Lc10, when “Lc15 [is] the point of invisibility for many users”?
Source: https://git.apcacontrast.com/documentation/APCA_in_a_Nutshell.html

QUESTION 3: Finally, why does the table go as high as Lc110 and Lc115, when the highest contrast is 106 (for black on a white background) and Lc107.9 (for white on a black background)?

Thank you kindly.
Here's an image of the mentioned lookup table with outlined squares using the noted widths next to the relevant Lc.

[Myndex]

Hi @angyb thank you for the questions.

_QUESTION 1: For accessibility, is it acceptable to use fractional pixel widths given that subpixel rendering can cause soft edges or visual blur, especially on lower-density displays? I would think we’d want to round to whole pixels to maintain sharpness and fidelity._



First, the CSS px is NOT pixels. px is the canonical length unit, part of the canvas abstraction layer, and is defined in terms of visual angle (1.278 arc minutes per px).

How a px is rendered to a given screen is a technology dependent function, and if it is anti-aliased or not, is completely dependent on the rasterizer.

But more, the effect of contrast reduction as a function of antialiasing is taken into account as part of the math and guidance.

QUESTION 2: Next, why does this table go as low as Lc5 and Lc10, when “Lc15 [is] the point of invisibility for many users”?

You'll notice that those levels are in red. And further, this specific chart you are showing is not part of the official guideline, it is experimental, and intended to generate the kinds of questions you are asking.

$L^c15$ is listed as the invisibility point for users with a profound visual impairment, that is, low vision up to statutory blindness.

QUESTION 3: Finally, why does the table go as high as Lc110 and Lc115, when the highest contrast is 106 (for black on a white background) and Lc107.9 (for white on a black background)?

106/108 are only the max $L^c$ values for sRGB and the public beta APCA.

The actual top is higher for some color spaces, and the actual top may be higher for APCA25, a newer version with direct CVD compensation. Again, these are shown here in red as "not part of guidelines" but an extension of the math/methods.

Here's an image of the mentioned lookup table with outlined squares using the noted widths next to the relevant Lc.

What you can take away here, is that each box should be roughly equal in visibility and roughly equal in perceived relative contrast, when viewed at the standard size/distance.

I hope this answers your questions please follow up if you have more.

[angyb]

@Myndex
Thank you for your kind reply. :)