What are the differences between the original (rasterized) CM fonts and the current (smooth PDF-friendly) type1 versions of these fonts?

It seems to me that the originals are much better looking. They are "thicker". You can clearly see the difference in printed output. The type1 ones are somewhat "thinner".

Why is this? Why aren't they exactly the same?

And final question: On a modern MacTeX distribution (on my Mac OS X), how can I compile my document with the older fonts?

Update: I feel the current answers haven't answered some of the sub-questions and, more importantly, haven't provided the proper understanding that I'm looking for. The question can be rephrased: I look at "Digital Typography" by Knuth and a recent mathematics book on my shelf. Both are clearly printed in very high resolution—the fonts are smooth and look how they are supposed to look. But the font in Knuth's book is far "thicker" and has a much warmer, juicier, and nicer feel to it. On the other hand, I recognize the font in the mathematics book as the modern ubiquitous "smooth type1" CM font. In comparison to the font in Knuth's book, it looks thin and pale. I'm looking for some foundational understanding for why this is the case and why it turned out this way. Any font experts in here?

  • 6
    It might be that MetaFont does a better job rasterising the fonts for small sizes than what PostScript raseterisers do (e.g. it might have some meta-info about the fonts that the PS outlines does not have). Mar 17, 2012 at 17:06
  • 1
    i don't think this really answers your question, but have a look at this answer that touches upon the same topic. it discusses the older font encoding's problems with hyphenation and copy/paste. it also references ot1 and t1 examples which made me think "wow, they are thicker!" Mar 20, 2012 at 3:56
  • 1
    Which resolution uses your printer? At which resolution were the PK fonts generated? Could you please include a scanned image of an actual printed sheet which shows the difference you stated?
    – JLDiaz
    Mar 22, 2012 at 16:44
  • Btw: The best rendering of CM I have ever seen was in "Beauty Is Our Business: A Birthday Salute to Edsger W. Dijkstra" - if you can, get the first edition somewhere. Apr 27, 2017 at 23:59

5 Answers 5


sorry, @Herbert, the image doesn't say it all.

what was neglected in the description is that metafont was run at a particular resolution to produce the first image, which, when scaled up, shows the artifacts of bitmapping.

here is a counterexample, adding in to your first test the type 1 cm fonts scaled in the "plain tex manner", along with the file that produced it. during the tex run, mf was launched automatically, with this report:

kpathsea: Running mktexpk --mfmode / --bdpi 600 --mag 1+0/600 --dpi 6000 cmr10
mktexpk: Running mf-nowin -progname=mf \mode:=ljfour; mag:=10+0/600; nonstopmode; input cmr10
This is METAFONT, Version 2.718281 (TeX Live 2010)

the output:

test of three cm/lm strings to compare shapes

that said, it's possible to see small differences -- the right-hand stem of the M and the stem of the R look thicker in the cm than in the lm. so i conclude that the type 1 cm fonts are more "true" to the original mf cm than are the lm fonts; nearly everyone these days except knuth uses the lm fonts because they're superior in many ways for work in languages other than english.

for folks who want to try this out for themselves without retyping, here's the input code; i didn't realize it wasn't possible to cut-and-paste from an embedded pdf image. (i included the code as "proof" that it was exactly what produced this output.)


\font\bigfont cmr10 at 100pt
\bigfont CMR\\[10pt]

  • Could the input code be part of the text of the answer, rather than an image? (I'd retype it myself, but I guess you've got the original available.)
    – Joseph Wright
    Mar 22, 2012 at 16:04
  • @Joseph -- i included it in the answer to "prove" that this was the code that produced the output. i didn't realize it wasn't possible to cut-and-paste from an embedded pdf. i've added the code, but left the duplicate in the pdf; hope nobody minds. Mar 22, 2012 at 16:12
  • @barbarabeeton: The pdf actually gets automatically converted to png when you upload it. Copying out of a png is not possible, of course.
    – doncherry
    Mar 22, 2012 at 16:23
  • @doncherry -- thanks. didn't realize this became a png insert. (some things i prefer to leave as "black boxes".) Mar 22, 2012 at 16:33
  • 2
    Not only MF can generate fonts at every resolution, but the raster files can be quite generally adaptable to each printing device. Something of this kind is available with hinted Type1 fonts, but not as generally as with MF.
    – egreg
    Mar 22, 2012 at 17:12

(As many readers tend not to follow links all the way through, and as links can die, this reproduces relevant bits of stuff referred to in the answer by bubba.)

I agree with the observation in the question. When I pick up one of Knuth's books from the library, the reading experience is different from when reading a typical book typeset with LaTeX. And the reason isn't just the style of writing, but includes typography.

Fortunately, there already exist high-resolution scans demonstrating this, by Raph Levien (who, incidentally, did this interview of Knuth, was himself interviewed for TUG, has extended the work on curves that Knuth and Hobby did for Metafont, and is a former maintainer of Ghostscript—knows a thing or two about font rendering).

Here is a 2400-dpi scan of a page from Digital Typography by Knuth:

Scan from Digital Typography

Here is a 2400-dpi scan of the output from a laser printer (HP LaserJet 1200):

Scan from laser printer

And for what it's worth, this is the "digital master": Digital master These images are from the webpage titled Effect of gain on appearance of Computer Modern:

Note the significant gain of the book sample, and slight weakening of the laser printed version, compared to the digital master.

Here are the three images again, at a smaller size:

From Knuth book From printer Digital

where you can see clearly even at this size that the hairlines are thicker in Knuth's book. So even though the modern printer output may be closer to the so-called "digital master", it's not what Knuth intended or designed for print. An explanation is on this page in German (translation via Google Translate, lightly edited):

Is Computer Modern really too thin?

Computer Modern is constantly accused of being too thin. Is that really true?

When D. E. Knuth created Computer Modern, the ink/toner spread slightly on the usual output devices (for example, the Xerox laser printer, which he used for testing, among other things), so that hairlines were slightly broadened. He took this into account, of course, and made the digital characters thinner than they should look printed.

For each output device, Metafont can create a matching font. You can specify different settings, including a value called "Blacker", which controls how strongly the letters are to be thickened.

The "blacker" value is, however, set as a default to low, which is why the Computer Modern is generally too thin for today's laser printers.

Regrettably, the authors of today's most commonly used Type-1 versions of Computer Modern (Blue Sky, CM-Super, Latin Modern) have used too low "black" values, so these fonts are too thin.

You can carry out your own experiments by picking up two books typeset in the “same” typeface Computer Modern, one by Knuth (where he'll have used bitmap fonts, tuned to the printer being used) and one by someone else (who will have probably used one of those Type-1 vector fonts), and scanning them at high resolution as above.

To answer your other question: if you want a similar output, you not only want to use bitmap fonts (make sure to set the resolution high enough considering your output device, of course, or you'll get junk as in Herbert's answer), but actually use them with an appropriate value of "blacker" for your output device and magnification. Examples of how to do this are in the answers to questions How to create new font which is thicker version of Computer Modern and Fatter Computer Modern (actually there are many questions on this site asking/complaining about CM being thin or “spindly”: of course they're all using the vector fonts). If your device is already one of those in modes.mf, you can probably use that directly. (The file was last updated in 2008 though, so it appears not many people are using bitmap fonts or adding their newer printers to that list.)

  • See also the figures in the article Computer Modern Roman fonts for ebooks by Martin Ruckert in TUGboat Volume 37 (2016), No. 3. Nov 4, 2018 at 1:42
  • Specifically, in the above article, Ruckert found good results with blacker set to 0.6, 1.6, and 2.4 for laptop (Dell Latitude E6530, 1920x1080 pixels, 142dpi), smartphone (Motorola Moto G, 1280x720 pixels, 329dpi), and ebook (Kindle Paperwhite 2, 1024x758 pixels, 212dpi), respectively. (Note that the dpi numbers are not directly comparable, because the typical viewing distance and the size in inches at which the text is displayed change.) Jan 12, 2019 at 23:00
  • See also packages newcomputermodern (using the "Book" weight instead of "Regular" weight: sample PDF) and mlmodern (sample PDF) which both provide versions of Computer Modern that are not too "thin". Jul 22, 2021 at 19:32

The image says it all ...





enter image description here

or the same with only cm fonts:





enter image description here

  • 3
    Whilst this shows the obvious difference between vector and raster, it doesn't explain the difference in print. As Enchilada notes, one would expect identical results on paper.
    – qubyte
    Mar 17, 2012 at 15:44
  • The above images are in original size! So its obvious that you didn't get the same when printed.
    – user2478
    Mar 17, 2012 at 15:49
  • 1
    And for smaller sizes? The question doesn't mention pixellated fonts. In fact, it suggests that they look better.
    – qubyte
    Mar 17, 2012 at 16:10
  • I have noticed a difference between T1 and IL2 encoding as well, IL2 seems to produce the "thicker" glyphs.
    – yo'
    Mar 17, 2012 at 16:33
  • 2
    For the non-vector fonts, the printer setting makes a big difference. Mar 17, 2012 at 17:00

Take a look at question (and the answers) Fatter Computer Modern. Basically, the reason is that, for a given outline, the Metafont rasterizer produces a "fatter" set of pixels than the current Postscript and PDF rasterizers. And, with Metafont, you can even tune the rasterization to the properties of the printer.


The original Metafont fonts are most likely better if they are used at their design size with the correct Metafont mode for the current output device: Metafont is probably the best rasterizer there is (it has been designed by DEK, after all). The type1 fonts may be better at other type sizes. But I doubt that many people still use the Metafont fonts; today we mostly create PDF.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.