Highest UPM fonts you’ve built? Shipped/sold?

John Butler

I’ve read a few threads over the years about unusually high UPM fonts and the challenges they create. The usual argument I remember in favor of high UPM is extremely thin lines for thin weights, high optical sizes or thin filigree ornament. (»Zier«.) And I understand 1000 is typical for cubic curve fonts and 2048 for quadratic. And I seem to recall 10,000 UPM was the working resolution for IK source files, but those are not final shipping fonts.

My question is what is the highest UPM font you’ve built, shipped and sold? And what limits or problems did it have, if any, in which particular applications?

Thomas Phinney

Here are some old threads on high-UPM fonts:
https://typedrawers.com/discussion/75/opentype-ps-with-2048-upm
https://typedrawers.com/discussion/123/maximum-upm-for-otf-fonts/p1

John Butler

Thanks, Thomas, for keeping track of those threads however you did, because I still can’t see them in the search results. I see that the OTF/CFF limit is around 4096.

My original question remains: who has done this, and built and shipped it? Who is selling a 4096 UPM font? Or higher?

John Hudson

We used 4096 a couple of times on Microsoft project, I recall, but for TTFs. For a long time, there were bugs in third-party PDF creators that presumed a 1000 UPM for CFF fonts, causing scaling and/or linespacing problems, so we avoided higher UPMs for that format. But the new v5.00 Brill fonts will have a UPM of 4000.

Thomas Phinney

Ben Blom has done and shipped a bunch of 3333 UPM fonts for OT-CFF. I do not know of anyone who has done higher than 4096 in either format.

James Puckett

Okay, I have to know, why 3333?

Igor Freiberger

@John Hudson Any technical reason for choosing 4000 instead of 4096?

John Hudson

The earlier versions of the Brill types were 1000 UPM, so I went to 4000 to give me more refinement but without introducing rounding errors.

Using power-of-two UPM was a Microsoft recommendation for TrueType when processor speeds were slower because it was slightly more efficient. I expect the improvement is barely measurable today, and Microsoft have been happy to let us use non-power-of-two UPM values when there was a good reason to do so (e.g. a late change of scale in Sitka).

John Butler

The prime factorization of 3333 is 3 × 11 × 101. My first guess would be something to do with width unitization, which often constrained and thus shaped older designs. 4000 still feels like a better choice.

Igor Freiberger

This is what Ben Blom explains about the 3333 UPM:

https://www.durotype.com/Info/3idx_Precision.htm
(Actually, not very much...)

Thomas Phinney

Ben Blom discusses their rationale for the 3333 UPM in the thread I linked earlier.
They actually design in a 10K UPM and drop the precision to ×⅓ later.

John Hudson

‘UltraPrecision™’ — I can’t tell from reading the explanation whether it is intended as satire, but for the record all my fonts go to 11.

Thomas Phinney

It feels like primarily a marketing thing.