W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > April to June 2017

Re: Mean character width for 16px Google Fonts (818 Families)

From: Wayne Dick <wayneedick@gmail.com>
Date: Tue, 30 May 2017 23:42:12 -0700
Message-ID: <CAJeQ8SDQD_fLSjXQqOMdPs=HEPZrLTVLu5cF9gLE=51wxLAkEw@mail.gmail.com>
To: Laura Carlson <laura.lee.carlson@gmail.com>
Cc: public-low-vision-a11y-tf <public-low-vision-a11y-tf@w3.org>, GLWAI Guidelines WG org <w3c-wai-gl@w3.org>
 *Font Width*

Two important issues came up during the discussion of personalization to
accommodate low vision. The first was line length. The group objected to
character count as a measure. The second was related to font substitution.
Would substitution of wide fonts disrupt layout severely. Since both are
serious from the user perspective, I decided we needed some more
understanding as to the distribution of font widths.

The Google Web Fonts database provides 818 font families covering most
categories of font. That seemed like a good starting point. I devised a
method to measure average space taken by one character when it is laid out
in a major browser. This is not the actual character width. It includes the
average amount of space each character is allocated by the browser to
separate it from other characters.

For my sample character set I used the ASCII visible character sequence.
That is characters 32 through 126. These are: (the space character, through
lots of punctuation, the digits, the lower-case alphabet and the upper-case
alphabet. I listed them in the order below:

!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_` abcdefghijk
lmnopqrstuvwxyz{|}~

I inserted the space between the ‘backward tick’ and the lower-case ‘a’ to
avoid any trimming by user agents. To promote string breaking I placed a
<wbr> between characters. That way the string could break anywhere. I call
this the Typewriter Set (TS). It is the keys you see on a standard
typewriter.
*The Algorithm*

Set font size to 16px. Place a TS in a paragraph element and give it a big
width. I chose 1600px, but 2048 would be safer. To measure the length of
the TS use a loop to decrease the width until the TS takes more than one
line. That is the width of the TS, W. The average character width is W /
(126-32)=W/94. I looped through the Google Font Families (regular variants)
and obtained some interesting results.

1.      The width of TS varies among browsers.

2.     The average character width in Chrome is 8.313px

3.  The standard deviation is 1.4px

4.   If we take the average of character width / the mean width you get
virtually 1. It is 1.000030982 precisely.

5.     591 font families lie within 1 sd from the mean (72%). That is close
to 2/3 the expected number for a normal distribution. We would expect
something pretty normal from a distribution of 818 averages.

Settles a few things.  For example, counting characters is a very practical
way to measure line length because fonts don’t vary that much. Also, 591
out of 818 fonts won’t disrupt layout more that changes in resolution,
screen size and other variables on the web. As far as needing to limit
choices for low vision to a small set of fonts. Well 591 will do, just in
the Google Font set.

A complete font width table is at.

http://nosetothepage.org/fontFamStatsChrome.html.



Sincerely, Wayne

On Tue, May 30, 2017 at 8:50 AM, Laura Carlson <laura.lee.carlson@gmail.com>
wrote:

> Hi Wayne,
>
> Thank you very much for doing this research. Based on this new
> information, do you still consider the spacing metrics in the Adapting
> Text SC [1] adequate for when a large font-family is selected and/or
> increased spacing is applied to text? It currently is:
>
> * line spacing (leading) to at least 1.5
> * letter spacing (tracking) to at least 0.12 em
> * word spacing to at least 0.16 em
>
> Do we need to increase it or is it still okay?
>
> Thanks again.
>
> Kindest Regards,
> Laura
>
> [1] https://github.com/w3c/wcag21/issues/78
>
> On 5/30/17, Wayne Dick <wayneedick@gmail.com> wrote:
> > I did some calculations.
> > It is pretty clear that the distribution of font width is pretty narrow.
> > I have more data but here is the raw stuff without analysis.
> >
> > http://nosetothepage.org/fontFamStatsChrome.html
> >
> > Wayne
> >
>
>
> --
> Laura L. Carlson
>
Received on Wednesday, 31 May 2017 06:43:27 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 21:08:13 UTC