kibo@jec311.its.rpi.edu (James 'Kibo' Parry) (04/19/91)
In article <91106.204118FLEGLEI@YaleVM.YCC.Yale.Edu> FLEGLEI@YaleVM.YCC.Yale.Edu writes: >I have to agree with the previous post (I lack an include prev msg function) >that Adobe is not so naive as to simply interpolate or extrapolate other >weights of the font. Having heard Sumner Stone, the former director of their >type development division, speak on how he developed the Stone families of >faces, it is obvious that they often use interpolations as starting points >for intermediate weights which are then modified into something appropriate. >For getting the stroke weight right, apparently, interpolation does >marvels; getting the feel right is another story. Yes--interpolation and extrapolation (with Ikarus or whatever) is a tremendously useful tool, but it is of course only a starting point, except in special cases were fonts were designed to be interpolated. Many fonts have characters whose design changes radically between weights (the "a" tail in Helvetica, the "e" and "o" of Times, the "R" of ITC Avant Garde Gothic), widths, etc. But for a font where you have drawn *TWO* weights that have similar characters, interpolation is wonderful for creating weights in between. Many commercial fonts have been created this way--including many ITC and Linotype ones. If the two ends of the spectrum are well-designed, and they "match", all the generated weights between them will look fine. The best way to spot a font where the weights were interpolated is to look at the samples of the five (or however many) weights in the catalog to see what shape the five samples make; if they make a trapezoid with a perfect diagonal at the right, that's a sign. (I suspect Monotype's Ellington as being generated that way.) Starting with *ONE* weight, of course, isn't good for much, no matter what font editor you use--you can thicken the strokes, slant the character, etc., but these effects just don't compare to actual bold, oblique, etc. (I don't believe anyone has ever done a good program to automatically make true oblique fonts from an upright font. Yes, there's a subtle difference between printing the regular font at a slant and having a real 'oblique'; if you merely slant an existing font, the vertical strokes get slightly thinner, and curves develop thick and thin spots.) It all comes down to the designer's intent; some fonts, such as Univers, were designed to be adaptable to different weights and styles, and others, such as Times, were designed to be used in a few specific weights or styles. In my opinion, the fonts that show the best design are those where all the styles *look* like they match, but in fact there are slight human-introduced variations to correct for various optical effects, etc. A good example is Futura (note how the proportions of the "o" change in the different weights, the tips of the "S" gradually flatten, etc., in addition to the major changes such as the apex of the "V" changing to a flat tip.) A related problem to the interpolation of weights, widths, etc., is the generation of different sizes from a single outline (where much of Metafont's emphasis is); some fonts are designed to be used at certain sizes, and at the wrong size the spacing might look odd, or corrections for optical effects might show up, and so on. Times and some imitation Futuras are notable fonts that had "small text" sizes designed. One font I'm currently working on is a roman font designed to be printed large at high resolution, because it has a lot of detail and subtle curves; one question I have to answer before I finish the design is whether I should also do a version with crisper corners, wider spacing, etc., for use at body text sizes, or whether the detailed version will scale down. (That's one problem with designing digital type--some people will be printing it at 3000dpi, and some will be printing it at 144dpi.) -- James "Kibo" Parry kibo@rpi.edu 132 Beacon St. #213, Boston, MA 02116 (617) 262-3922