Vertical Metrics or Scaling within the same UPM and Printed Appearance

Jacob Casal

The last paragraph holds the question, and the preceding ones hold background information leading to it.

I have been a bit reluctant to ask this question as I feel that the answer I’m missing is going to be blatantly obvious, yet I can’t seem to wrap my head about how to go forward. I will briefly note I have been reading the various font editors’ guides (Fontlab, Glyphs), design resources (Karsten Luecke, Font Development Best Practices), and many threads here on this.

I’ve been digitizing an early Didot (around 1790–1800), and I was a real greenhorn when I started this project. I had researched type from that period as well as modern Didots (LT, HTF, Theano) and chose 1000 UPM while emphasizing grids of 88 units and smaller spans of 8 units (e.g. a serif being 16 units tall). Historically, (e.g. grids for the Romain du Roi) 8 made sense; many lengths of modern Didots seemed close to 8 too. Units of 8 has been working beautifully between my design’s accuracy to the manuscripts.

However, I got better at measuring letters from the manuscripts and noticed how my design seemed off. The key manuscripts are in 15 didot points (16 points), which is key to my confusion. When scaled down to match the x-height, I realized my ascenders and descenders were far too short (40 units away from an /l or /p; 56 units for /f or /y as it has deep overshoots ). Further, scaling down LT Didot’s x-height (shorter than mine, I made sure it was scaled in terms of its own design) to the manuscript’s was also too short. Currently, my design matches the manuscript as I wanted it to be accurate.

With all that said, my confusion lies in metrics or scale. It seems I could keep it as is, and with some finagling of the metrics still be functional. However, if it were printed at 16 points (physically around 0.2222in/5.6444mm) then it would, point size to actual size relativity across digital fonts aside, be bigger than the manuscripts 16 points since my ascenders and descenders go past the 1000 font units the program is scaling down. In scaling down so that ascenders and descenders all fit within the 1000 font units it loses that whole units around 8 thing and becomes a bit less clean. Despite looking smaller at 1000 font units would it better match the measurements of the manuscripts? Am I better off continuing as I am or scaling it down to best match how it would have looked if printed in the manuscripts. Part of the scaling concern comes from the whole relativity of point sizes across various fonts, but is it the best option?

I feel as though I am sorely wrong and obtuse in my reasoning and that the answer is very simple, but I haven’t been able to put my finger on it.

Georg Seifert

The pink bar should be as high as the UPM. What you can try is to increase the UPM value a bit, export, test, repeat until you are happy (or, calculate the value from measurement)

What ever is as long as the UPM will measure to the set point size. Find a measurement (like ascender + descender) where you know how big it should be in print, print it (or export to PDF), measure it and the discrepancy is the factor you multiply your UPM with. And just change the UPM, don’t use the rescale function. 

When you happy with the resulting size, you might want to scale the font back to 1000 units.

Jacob Casal

Many thanks Georg and Thomas, I’ll give it a go. As I thought, I was overthinking it. Just running through your instructions to make sure I’m doing it right:

Pink bar currently represents 1000 UPM, so for now change it to match the total units of asc. + desc. (56(2) + 1000 = 1112 UPM). This UPM when scaled down will match the 16 point size (or any other point size set) of print. As you say, changing the UPM has the same effect as a scaling function, so no need to scale; despite seeming smaller than the other Didots, it indeed fits when sizing the manuscript image to 16 points!

The only part I’m not clear on is when you say to scale the font back to a 1000 UPM. Wouldn’t this undo the adjustments?

Jacob Casal

On thinking about the last point further, you were just reiterating letting the change back to 1000 UPM handle the scaling, yes? Since 1112 is an odd UPM to have, I would then have to make the necessary adjustments to get rid of any fractional anchors back at 1000 UPM, correct? Thanks again, I can finally start making more progress in the designs!

Georg Seifert

You got it.