Try Google “pixel font”.

> I am looking for some really diverse sans serif fonts (extra light, light, condensed, roman, bold, caps etc…) that are legible at small sizes on the screen (7-9 points).

This is hard (some would say impossible) - unless you have ClearType (or one of its clones) activated the best you can do with bi-level rendering is have one or two pixel stems which means a maximum of two weights.

7, is that seven? Even if you have CT, I doubt it’ll help. 8 a little. 9 a little more. Stylistic diversity at those sizes is not a problem because it’s not really possible. Have you ever done this before? :)

Perhaps it could be alphabetic diversity, eg unicase.

First some groundwork: to me the best way to
measure the size of a bitmap font is to go from
the top of the lc ascenders to the bottom of the
descenders. Any other measure is less useful
in the end.

—

Guys, it’s not so bad. Certainly no need to resort to unicase.
And you don’t need ClearType either. In fact at those sizes
CT’s deficiencies (specifically color fringing, inconsistent
glyphs) get amplified.

Topologically, you need five pixels for the x-height.
At 9, you have no real problems. At 8, the descenders
get tricky. At 7, you’re left with only two pixels for
both ascenders and descenders, which makes shapes like
the “f” and “g” (in either form) impossible. HOWEVER:

1) Just like the masters of phone book font design
have always done, you don’t necessarily have to respect
the baseline and x-line all of the time: for example the
bar of the “f” could be lowered, and the head of the “g”
could sit higher than the baseline.

2) You can actually use 4 pixels for the x-height,
although easy decipherment suffers a good deal.

My Mana family (a grayscale pixelfont, what I consider
as having the highest aesthetic* and functional potential)
comes in 16, 13 and 11 PPEM sizes. I have a 9 planned.
It wouldn’t be very hard to do an 8 (and even a 7) but
I won’t; but only because at those sizes it’s no longer
Mana, and becomes too close to free stuff that looks
almost the same.

* Yes, including stylistic differences.

If/when I make pixelfonts smaller than 9, they will
have to have something quite different about them.
(Even more different than using gray pixels.)

hhp

Can you post something? intrigued how you can get an extra light, light, medium, regular, bold, extra bold, black and condensed with 9px to play with.

Si

You vary weights sort of in the same way that “blind” automatic-from-outlines algorithms do, just with a lot more control (stability): you use shades of gray.

In terms of the black bodies, the threshold for going from one- to two-pixel stems varies according to: PPEM size; x-height; and width. For a design of conventional Cartesian proportions, the threshold is about 14-15 PPEM.* Once the black body thickness threshold is established, you can use a dark gray at just above the threshold to get a demi weight for example. This, I’ve done myself, and it produces quite nice dark-weight emphasis (much nicer than a full-bold and even an italic). On the other hand I admit I haven’t tried Light weights that way. But it’s nothing a commission can’t fix. :-)

* It’s mainly with focus-group-induced mayhem that the original MS core fonts started violating this and rendering out too-light one-pixel bitmaps at “normal” reading sizes. And Apple followed suit of course. Before the MS core fonts, Apple’s 16 PPEM sizes had nice 2-pixel stems.

One thing that helps btw is making the forms wider the smaller the point size. Although this can skew the overall character of the family, it: leaves more room for small[er] variations; and correlates nicely with optical scaling needs.

hhp