laser-lovers@uw-beaver (01/10/84)
From MACKAY@WASHINGTON.ARPA Mon Jan 9 16:45:22 1984
We have just achieved what we fervently hope is a fairly stable version
of our DVI2LGP driver for the Symbolics Inc. machine. I am a bit
at a loss what to say about the quality, since it is obvious that
our model is defective in several ways. For a starter, it swallows
up single-pixel vertical lines when they are placed at certain positions
(probably at either the right or left hand end of a byte or sixteen-bit
boundary and the boundary coincides with some other arithmetic value.)
The result is very evident in things like square-brackets, which thin out
abruptly at certain predictable positions in the page. That seems quite
certainly to be a controller problem, and not a problem with the Canon
engine proper.
The application of toner to paper seems fairly haphazard---what I
have come to know as ``spitball'' mode. The letters are much coarser
in definition than similar letters on our old Versatec 1200A. They have
furry edges, and are very much blacker, as if spattered onto the page.
We use raw METAFONT products for our font library, but we are
quite aware that any font at this sort of resolution ought to be
bit tuned. The Computer Modern fonts are presently changing often
enough that we are disinclined to get started on the irreversible
job of bit-tuning. There is a good pixel-editing program developed here,
(Tops20 only at the moment, but it is a WEB file and hence more
nearly portable than a lot of software) which Lynn Ruggles has used
with success. It is interactive, with Ansi X3.64 escape sequences
used for cursor positioning. That means a rather distorted image on the
screen, since character cells are nearly always off-square, but it
would be possible to work something out for a bit-mapped display.
There is no doubt that when Computer Modern stabilizes, pixel-editing
will improve it at this resolution, but we would be hard put to it to
justify the effort unless we get a better controller.
I think there would be little opposition to the claim that
Computer Modern has been vastly improved over the past year. But they will
never be atg their best at low resolutions. Genuine low-resolution
fonts for office automation equipment raise specific problems of
design (as Brian Reid has most vigorously pointed out), and a low-res
analog of a font that is really meant for 1800 line/inch devices or
better will always suffer by comparison. The letter-spacing
can get quite jumpy when a high-resolution font is cut down to
something like the .1200PXL fonts that we run on our Symbolics Machine.
I use the following text as a display sample with font charts
produced on either the Versatec or the Canon machine.
The design of low-resolution typefaces is complicated by the need
to achieve two different and rather incompatible ends. A font which is
exclusively intended for Office Automation equipment should be designed
to make the very best use of that environment. Its proportions, balance,
shading and letter-spacing can and indeed must be fitted to exact
pixel-boundaries, whether or not the refinements of anti-aliasing are
used.
A low-resolution font which was created primarily as the proof-copy
imitation of an associated high-resolution font will never be quite
so precisely matched to its environment. Its contours will necessarily be
determined by exact pixel-boundaries, but often not in the way we might
prefer. It need not be sloppy, but it can never be so well tuned as
a font of the first variety. It is a poor Platonic reflection of an
idea that cannot be fully known except in the dazzling light
of high-resolution typesetting. The vast majority of fonts produced
with METAFONT belong to this second variety. A low-resolution
METAFONT character should be judged for what it is, and not for what
it has never claimed to be.
Pierre MacKay
-------