Testing mftrace

I am not entirely sure how fair of a comparison I am making, but I am open to suggestions on how to better compare the hypothetical output of a font designed using METATYPE1, where a Type 1 font would be directly generated, versus a font designed using METAFONT, where a Type 1 font is generated using mftrace.

For my experiment, I chose to compare glyphs from the Type 1 version of Computer Modern maintained by the American Mathematical Society, which was presumable crafted by manually tracing the bitmap version and optimizing various aspects by hand, with glyphs generated by mftrace on 3000DPI bitmaps generated from the METAFONT source. I could imagine that it might be fairer to compare the result of using nearly identical METATYPE1 and METAFONT source to generate the Type 1 glyphs, but then I would be biasing the design process towards the limitations of METATYPE1. I had considered using the METAFONT source for AMS Euler, because the source is simply the outline and would be easy to convert to METATYPE1.

I also decided to use glyphs with plenty of curves for the test, as I figure that tracing software can probably do a pretty good job with straight lines.

In any event, I will leave it to you to decide whether you can distinguish which is version is which below.  The order in which the two versions appear differs for each of the three examples below.

qaqb

sasb

ampaampb

As you might have guessed, my opinion is that results are so nearly indistinguishable, that given the design limitations of METATYPE1,  it would make much more sense to work with METAFONT and mftrace, and use FontForge or a similar tool to add hinting as a postprocessing step.

2 Comments »

  1. Samuel Bronson said,

    February 1, 2011 @ 7:29 pm

    Hmm, very nice, but how many points does each outline have?

    Also, did you notice that the tail of the Q on the left curls back on itself more than the one on the right? At least, there’s a critical point on the right side of its tail that’s rather further down than on the right-hand Q, which doesn’t occur until the nearly-circular curve at the end of the tail. So, I’m going to guess that the left-hand Q is the one generated by mftracing Knuth’s very own Q. (I might be giving mftrace a more credit than it’s due ;-)

  2. Shriramana Sharma said,

    May 6, 2012 @ 11:41 pm

    Hello. I came across this very interesting post while googling for MetaType1.

    I find interesting your contention that metafont + mftrace @ 3000 dpi is much more productive than MetaType1. I have been investigating the best way of producing a good font (see http://tug.org/pipermail/metapost/2012-March/002495.html) using the Meta* tools and would like to know more about your experiences with mftrace. As I don’t find your email ID on this website, could you please drop me a mail to my GMail ID (which I have input in the appropriate field in this form) so I can learn somethings from you?

    Thank you very much!

RSS feed for comments on this post · TrackBack URI

Leave a Comment