First question:  How exactly does the EM relate to the actual glyphs?
For example, let's say I have a sample in print that I want to make
into a font.  I don't know what point size it originally was, but can
make physical measurements.  If I measured the vertical distance
between a typical descender and a typical caps height, would that be an
appropriate EM size?  What if the sample has no descenders, such as
small-caps handwriting?  I ask this because, historically, I have had
trouble getting the size of my glyphs and the EM size to be in the
right ratio that the text size is comparable to other fonts at the same
"point size".

Second question.  I know the reason why TTF fonts have EMs that are an
exponent of 2 when measured in f-units, but how problematic would it be
if it wasn't?

--
Vid the Kid

VidTheKid wrote:
> First question:  How exactly does the EM relate to the actual glyphs?
> For example, let's say I have a sample in print that I want to make
> into a font.  I don't know what point size it originally was, but can
> make physical measurements.  If I measured the vertical distance
> between a typical descender and a typical caps height, would that be an
> appropriate EM size?  What if the sample has no descenders, such as
> small-caps handwriting?  I ask this because, historically, I have had
> trouble getting the size of my glyphs and the EM size to be in the
> right ratio that the text size is comparable to other fonts at the same
> "point size".
> 
> Second question.  I know the reason why TTF fonts have EMs that are an
> exponent of 2 when measured in f-units, but how problematic would it be
> if it wasn't?
> 
> --
> Vid the Kid
> 


VtK,

An em is nominally a square that is the same dimension as the point size 
of the glyph. If you compress a font in the composition software, you 
squeeze the em into a rectangle narrower than it is high. If you expand 
it, you make it wider than high. However this has NOTHING to do with the 
nominal squareness of it internally to the font--it's just an ugly 
scaling trick that respectable people rarely use.

Maybe Character can address whether the em is still a square in a 
condensed or expanded font. To my understanding it is, regardless of the 
width of th M glyph.

So the fact that it is however many units wide it is in TrueType or in 
Type 1 is just a given that you have to work with. That doesn't keep you 
from defining other fixed spaces at whatever width you feel is appropriate.

Now, back to the first question. If you have a multiline sample of 
running text that appears "normal" to you in terms of leading, or that 
you know, for example, was created in a word processor like MS Word or 
in a typical layout program such as PageMaker, then what you want to do 
is measure baseline-to-baseline between two lines and divide by 1.2. So, 
for example, if two lines measure 14.4 pts b-to-b (not an easy amount to 
measure precisely, but if you measure 10 lines and divide by 10 you'll 
get a good number), then divide by 1.2, you get 12, which is the size of 
the type. This is because the default leading in most programs is 120% 
of the point size.

If the leading is not standard or if you only have a single-line sample, 
then measure, as you suggested, from the bottom of a descender to the 
top of--not a cap--an ascender. In many fonts, the caps are a bit 
shorter than the ascenders.

Beyond that, it's a bit of guesswork, as you really can't measure 
precisely from a printed sample.

Good luck,

Dick

Dick Margulis wrote:
> An em is nominally a square that is the same dimension as the point size
> of the glyph. ...

Yes, I know that the EM, when scaled with the text, is what is measured
to give the text's point size.  What I really wanted to know was, what
size to make the EM compared to the actual glyphs -- or, as the case
may be, how large to make the glyphs compared to the EM -- when
designing a font either from scratch, or from an abstract sample which
might have non-standard leading or only one line of copy.

> then measure, as you suggested, from the bottom of a descender to the
> top of--not a cap--an ascender. In many fonts, the caps are a bit

I suppose that's a good guideline, but is there such a guideline that
covers fonts that don't have descenders, such as all-caps or small-caps
styles, besides experimenting and going with what "looks right" ?
Thanks anyway for the partial answer.

And I still would like to know if a TrueType font having an EM size of,
for example, 1536 FUnits, would likely present significant technical
difficulties in any programs, operating systems, or printers...

--
Vid the Kid

> Yes, I know that the EM, when scaled with the text, is what is measured
> to give the text's point size.  What I really wanted to know was, what
> size to make the EM compared to the actual glyphs -- or, as the case

There is no exact relation between glyph shape and em.

In the age of metal type it was easy to measure em. It was the same as the height of the type. Em space, of course, was a rectangular non-printing piece of metal.

Em usually is the distance from ascenrdes to descenders but not exactly. Both have to fit into em height but usually there is a small gap so ascenders and descenders will not touch with 'default' line spacing is used.

Basic rule is that letter M should fit into em, hence the name. The rule, luckily, is hardly ever followed.

There have been several attempts to standardize type size and bind the glyph size to point size. They have all failed miserably. Good typefaces just cannot be made mathemaically and by measuring and stanrdizing. If you study closely typefaces like Futura you'll notice that there a lot of little non-geomertric tweaks that make the typeface look more geometric.

Jukka

VidTheKid wrote:
> <snip>
> 
> And I still would like to know if a TrueType font having an EM size of,
> for example, 1536 FUnits, would likely present significant technical
> difficulties in any programs, operating systems, or printers...

the short answer is No.

there was a discussion on this in the FontLab forum, I believe, but I 
can't find it  --  possibly becuase it was under a different thread heading

Type 1 uses a grid 1000 units high, and as wide as you want to make it

you will find recommendations that TrueType be drawn on a grid 2048 tall

ignore them

the recommendation on Fontographer's old site, was that the height of 
the grid should be a power of 2  --  I took a 1000 unit Type 1 font, 
redfined the grid height as 1024, and the resultant TrueType font had a 
little extra leading built in, but worked fine

in fact, had I been a bit braver, I could have left the grid height at 
1000, and it would still have worked fine

if I recall the FontLab discussion correctly, timing tests showed no 
advantages for powers of 2, and it was suggested that this powers-of-2 
biz was to take advantage of some bit-stuffing techniques for faster 
division on the 680x0 used in early Apples

see, what happens is that a font's outlines are scaled down to one point 
in size  --  that's where the division takes place  --  then, when you 
specify a new point setting, those 1 point outlines are scaled up  -- 
which is multiplication, and a whole lot quicker

so, if 1536 is the number that does it for you, then go with it

if you feel the need for something exotic, why not try a prime number? 
or a power of 60, for lots of divisibility?

all the best   . . .   /phil