Kent Lew

I don’t necessarily agree with this. Depending upon the relationship between styles, I have found that I can often start from previous kerning. 

If I am methodical and consistent in my development of a heavy weight, in fact, many pairs can share the same value across fonts, and other pairs may take a common adjustment.

For example, I just quickly analyzed a recent Ext Latin sans-serif family I delivered, and the Ultra pole shares about 80% of the kern pairs/values with the Regular. Another ~15% are the same pair but with adjusted values, and the remaining ~5% are pairs unique to the Ultra. (There are also a few pairs unique to the Regular.)

This will not always be possible, of course. And not to the same degree with all styles of type. But neither is it inconceivable.

There is no fixed formula, but there can be rational relationships and, with experience, one can be methodical about it all.

And, yes, the eye should be the final arbiter.

This was a driving principle behind the Standard Lining (aka Point-Line) system that was being adopted by founders in the U.S. in the 1890s and elsewhere in the first decade of the 1900s, not unique to the German implementation.

(I don’t think Dan meant to imply as much; I just thought I’d clarify for anyone who is unfamiliar with this history and who might have read it that way.)

Here’s an illustration of the scheme.

Or December.

Fair enough. First week or so can work. General wisdom and experience at Font Bureau, though, was that December releases were disadvantaged. We might release an expansion, because that was often aimed more at existing customers who were predisposed to pay attention. But not something new and exciting.

Of course, you can’t just lower the heights of your lowercase letters and expect proportions to still look right. Everything will start to look expanded.

Unfortunately, the kind of late-stage surgery you’re contemplating can be non-trivial.

For a starting point, you can try something like this:

Figure out what percentage you would need to scale down your lowercase to achieve the x-height you think you want. Then calculate the increased stem width you’d need to start with in order to end up with your current stem weight after scaling. Then calculate an extrapolation percentage between your Thin and Heavy masters that will get you to this extra heavy weight, generate that instance, and scale down to your target.

For example: say your Heavy weight has a lowercase stem of 200 and you want to scale down 90% to adjust your xheight. You‘d need to extrapolate an instance with 200 ÷ 0.9 = 222 stem weight. Say your Thin weight has a lowercase stem of 20. The difference between your Thin and Heavy stems is 180. To get to 222 stem weight, you need to add 22 units as a percentage of overall difference: 22/180 = 0.122; +1.0 for extrapolation = 1.122 = 112.2%.

In this example, you need to extrapolate 112% from Thin to Heavy, then scale 90%. Of course, you’ll then need to review, refine, and correct, but you might find yourself with a reasonable starting point for the redraw.

(With such a large starting differential in xht between your Thin and Heavy masters, you will end up with a slightly increased xht in your extrapolation, which will affect your scaling, and if the percentages are significant, you may need to compensate in your calculations.)

Of course, you can also find UFO tools that offer just this sort of multi-dimensional scaling/interpolating with an interactive UI — like UFO Stretch, for instance. Really, what you’re looking at is not conceptually different from deriving small caps — scaling while maintaining proportion and stem weight.

Good luck.