Bravo!

I can confirm my sample is a photocopy. Unfortunately, neither any original text nor the typewriter itself are still in my possession. :-(

Nice project! I've started several times to digitize a typewriter face that I'm in "love" with, and still am not sure about the best way to go ahead. By now I'm starting to think just to look at the daisywheel with a loupe and drawing everything by eye may be way to go. I just couldn't make scans of typewritten letters that showed enough detail in the letterforms.

Before you start drawing, it may be useful to do some math to save you trouble later.

Typewriters had a fixed spacing (a certain number of characters per inch). You can find it out like this:

Type the same letter, for example I, about twenty times in a row. then measure from the stem of the first I to the stem of the eleventh I. If the distance is 25.4 mm (= 1 inch), then the spacing would be 10 characters per inch. If 13 I make up 1 inch, it would be 12 characters per inch. If you measure multiples of 1 inch, and divide accordingly, you will get more accurate results.

There were a couple of models, especially in Europe, which use the metric system and would space 10 or 12 characters per 26 mm (instead of 25.4 mm).

If you draw your font on a 1000-unit grid and make the characters 600 units wide, you will get exactly 10 characters per inch at a font size of 12 points, or 12 characters per inch at a font size of 10 points.

If you have a metric typewriter with 2.6 mm spacing, you will need 614(.17) units character width.

Good luck :)

There were a couple of models, especially in Europe, which use the metric system and would space 10 or 12 characters per 26 mm (instead of 25.4 mm).

To be precise, the 26 mm applied only to the 10-per-inch size (2.6 mm ≈ ⅟₁₀ inch). For the 12-per-inch size, the 2.1166 mm ( ⅟₁₂ inch) would be rounded to a near number, but I'm not sure if it was 2.12 or even 2.2 right now and can't find my typewriter catalogue :(

That is helpful advice: the typewriter I used has exactly 12 characters per inch, and 600 units (out of 1000) indeed makes the font have exactly the right size at 10 points. (Alternatively, I suppose I could use a width of 500 units, which would make the right size at 12 points. But 600 units seems like the best choice when compared to the size of other fonts on my system.)

The line spacing of the machine was 170.5 mm per 40 lines, which works out to 12.08 pt per line. (It is notably not 12.1 pt, which would make 40 lines span over 170.75 mm; my measurement of 170.5 mm is, if anything, rounded up rather than down. But I don't know if 12.08 is the exact space intended by the manufacturer, or if it really should be 12.1 pt, and this particular machine is advancing the paper slightly less than intended?) In any case, if I have understood it correctly, there is really no good way to control that as a font designer? I notice that my test fonts (i.e. the one with 600 units @ 10pt and the one with 500 units @ 12pt) both have the same large line spacing of 14.1 pt (in the first and only program I tested in, OpenOffice). So, to get the correct line spacing, I suppose I simply have to instruct the users to set it explicitly in their typesetting program. Or is there a better way?

What you write about inspecting the daisywheel with a loupe is intriguing, but for me as an amateur it seems like a daunting task: I wouldn't trust myself to be able to reproduce the proportions of the letters correctly. However, for me personally, it is a moot point: I don't have access to the daisywheel. I only have the typed pages, which were made on location in a shop specializing in repairing typewriters. I think, though, that by generating an average picture of several specimens, I will be able to get close enough (for my purposes) to the actual shape of the types on the wheel. (But it would indeed be very interesting to investigate that further; consider for example the small indenture on "s" which I mentioned above, and which I presume is a small damage on the type. It would be great to be able to confirm that by visual inspection of the wheel...)

I have uploaded a first draft.

I have here mainly concentrated on getting the letters uniform, making the stems consistently wide and reusing curves which should be identical (the right part of "e" is copied from "o", while the right part is identical to that of "b", and so on). The main difficulty I have faced is to decide to what degree I should simplify the letters. Primarily, the stems of "bdfhknmpqtu" have a noticeable swelling in the bitmap picture, which I have not reproduced. Thad however might have been a mistake; especially my "f" gives a very different impression, I think. So I will probably reconsider that.

Another thing to notice is that I have deliberately drawn the horizontal elements of "ilrz" slightly at an angle. This is based on the scans, but again, that is something I have my doubts over.

Anything else I should think of as I proceed?

I can now present a new draft. As can be seen in the PDF, I have now drawn glyphs for all characters present on the typewriter. Now, are there any things in particular that you think that I should work more on?

There is one aspect in particular that I am unsure of, namely how corners should be treated. There will naturally be some distortion of the letters when they are written on the typewriter, leading to sharp corners being more rounded, and the averaging process that I have used (mentioned above) may also introduce some further smoothing. To what degree should I adapt to this in a re-creation? Right now, I have made a compromise: angles less than 180° (measured on the inside of the contour) have been rounded, but angles larger than 180° have not, except for when they approach 360° (i.e. ink traps). Is this a reasonable approach, or does it seem inconsistent?

W,

I'm not enough of a typographer to answer your specific questions, but I will say that it is looking very good and true to the original design to my amateur eye. :-)

winge,

I think Calibri can be a proof that your treatment should work.

After having been distracted by other things, and letting this project lie dormant for close to one and a half years, I can happily announce that the font is finally finished, and available for purchase from MyFonts, under the name Cubiculum:
http://www.myfonts.com/fonts/johan-winge/cubiculum/

Many thanks to everyone helping me in this endeavour!