Proper weight instance progression for a multiple master

Andrea T.

Hello everybody,

I just want to know if there are articles, or at least discussions, where i can find solutions for a proper instance weight progression for a multiple master font.
Now i'm tryin to use impallari progression, but...Well, i don't understand so much how to use it  

Thanks a lot

Chris Lozos

There was a discussion about this a few months back.  It also included Lucas's method. http://typedrawers.com/discussion/comment/23208/#Comment_23208

[Deleted User]

The user and all related content has been deleted.

AbrahamLee

I've been thinking about this a lot lately and discovered something interesting. The CSS font-weight# progression (at least how they're described here) follows the nearly exact opposite Ogee curve from the Impallari progression:



The best-fit Ogee curve of this progression is given by the purple dashed line while the true CSS values are the connected blue dots.

NOTE: The exact names/values in this extended set of CSS values (or whether the values outside of the round 100s are even valid) can be debated, but that's not important. The purple Ogee curve is the important part.

Here's my take on the weight progression: Does each weight look different enough from both its neighboring weights? I found that the Impallari progression means too little differentiation on both light and heavy ends of the weight spectrum. Why? Take a look at this chart:



The 1-18 index refers to the CSS weight designations above, of course. That's what each progression looks like, but if that's too abstract for you, try this set of graphs instead which represent the relative weight (as if they were stem thickness, for example):



Have a good long look at that. Notice anything? We can talk ratios all day, but here are some general observations:

- Linear: Lighter weights are more distinct, gradually becoming less distinct as the weight increases.
- Impallari: Lighter weights are less distinct, middle weights are more distinct, heavier weights are less distinct.
- CSS: Lighter weights are more distinct, middle weights are less distinct, heavier weights are more distinct.
- Lucas: Lighter weights are less distinct, gradually becoming more distinct as the weight increases.

I tend to like the CSS progression best because most of the spectrum ends up being distinct from its neighbors. I also like that the body text weights (Book/350, Text/375, Normal/400, Thick/425, and ExtraThick/450 on these charts) are different enough, but similar enough to offer "grades" to work with, depending on the target medium.

The golden question: What's the right approach? There are all sorts of progressions one can try. As has kind of been the consensus already, you're going to have to eyeball it . The above observations can at least guide you depending on how you'd like the progression to pan out.

My two cents.

P.S. I think it would be interesting to survey some of the super families and measure their stem thicknesses to see how their weights progress.

Christian Thalmann

Abraham, I think the plot with the stem widths lined up is a bit misleading. One of the big reasons why the question of weight progression is non-trivial is because for heavy weights, the size of counters and spaces becomes the dominant visual characteristic. If your heaviest weights are placed relatively far apart, the counters will seemingly collapse abruptly in the last weight.

Chris Lozos

What is grown from a proper mathematical progression or a smooth visual progression are not necessarily the most useful variations of a font weight.  As we approach the extreme heavy weights, the number of visual problems increase greatly. As Christian stated above, the counter space becomes critical.  With it, the letter spacing requirements get more problematic.  A low contrast sans serif has the most difficulty at this extreme. Higher contrast fonts give more flexibility to solve visual problems. A type designer's job is to make a family of fonts useful together, not fit a curve.  That is not to say that both cannot be done. That is to say that fitting the curve is far less important and should not occur at the detriment of the families' usability for the job intended.

AbrahamLee

FWIW, I completely agree with both of you. I'm certainly not professing to understand this perfectly and very much respect your point of view based on experience. So, thanks for your feedback on my analysis! (BTW, I hold both @PabloImpallari and @Lucas de Groot and their expertise in the field in the highest regards.) 

Design trade-offs will likely need to be made to make a specific interpolated weight work right. Weight progressions may not (ever?) fall precisely on any given curve. Still, I think it would be interesting to see how some of the "industrial strength" super families have decided to progress their weights and (perhaps unanswerably) why. This would probably be the most insightful study of all. But who has time for that kind of thing, right?

I might just do this myself. Any recommendations on respected font super families with 5+ weights?

Peter Constable

Just a quick note: in an OpenType variable font, you can use an 'avar' table to tweak a weight progression when interpolating between two "masters".

PabloImpallari

@Abraham Lee I love when people keep thinking about a problem, make exploration and find new ideas and solution. This inverted ogee does make a lot on sense.
It will be nice to have it displayed in the calculator, so I will try to add it when I have time (next month or so). I will call it the "Abraham" curve, since CSS means a lot of other things.

AbrahamLee

PabloImpallari said:

@Abraham Lee I love when people keep thinking about a problem, make exploration and find new ideas and solution. This inverted ogee does make a lot on sense.
It will be nice to have it displayed in the calculator, so I will try to add it when I have time (next month or so). I will call it the "Abraham" curve, since CSS means a lot of other things.

Ha! Sounds good.

FYI, what I found from the CSS data is that it's really a weighted Ogee that pushes it *slightly* towards the linear progression, like this:

Ogee(x) = Linear(x)*(1-t^1.25) + Power(x)*(t^1.25)

where t = (x - xmin)/(xmax-xmin)

Lots of ways to formulate it!