Good post Brian! Too bad no-one's replying...

Some years ago I read somewhere that Luc(as) de Groot who designed TheSis had some mathematical formula to determine weights and the differences between them. In short he said that the optical average between light and bold is *not* 50%. I can't remember where I read it but I'll keep looking :-)

http://www.amazon.com/modification-letterforms-Stanley-Hess/dp/0910158037

Most type designers are too tentative in varying stem width for optical balance.
Not Frutiger.

And don't be afraid to make a reeeeeally bold Ultra weight, even without an italic.
As I've said before, design your Hairline and Ultra early on, and interpolate between them for the other weights.

So what you are saying is
a.
Don't be afraid to vary stem width for optical balance. (Frutiger is balanced at that), and
b.
Make your reeeeally bold Ultra weight and your hairline weights early on, and interpolate between for your other weights

And on this, I have heard that it's a bad idea to interpolate between very bold and very thin weights, because they are so extreme. Would it be better to interpolate between regular and these other weights, or directly between these?

Does anyone have any examples of this?

And also I'd be interested in seeing that Luc(as) de Groot quote.

http://www.google.com/search?q=groot%20interpolation%20site:typophile.com

I usually interpolate between Hairline and Regular, and between Regular and Ultra.
Although on some occasions I have had to go back and do another Ultra, because I didn't go dark enough the first time.

Ok, thanks!

Also, here is an article explaining how the golden section may have something to do with weights, and explains de Groot's formula briefly. There are more links in that thread. Gems!
http://typophile.com/node/32313

also this one for shifting vertical metrics for darker weights (this applies in reverse to light weights)
http://typophile.com/node/65057

Mr. Shinn, when you said not to be afraid to make a "reeeeeally bold Ultra weight", was this what you meant? I just saw a diagram on de Groot's website showing a bold stem the width of five stems, and thought, why not? I'm wondering how well this will interpolate...

500% bolder? I dub this the "Way Too Bold" weight.

...was this what you meant?

A young Gilles Villeneuve spun out at every turn during practice, laughed at by the journeymen drivers, then won the race.
His theory, put into practice: How does one know how fast one can corner, if one doesn't know how fast is too fast?

How do you know it won't interpolate unless you try it?

Haha!

Having tried it, the answer is: this fast is just a tad too fast!

Interpolated with the regular
0%

5%

10%

20%

40%

80%

100%

The problem is, with letters like s, a, and e, the structure and thickness has to be changed too much, and so the intermediate weights suffer. Probably, a little above what is 50% interpolated in this test would be about the boldest I can feasibly go and have it interpolate well. I could, of course, make another master, and go all the way, but I can't go the far with just on bold master.

I feel I have overcome something great. I also learned that you have to make the starting point the same on each glyph.

I am not a type designer, so don't take too serious what I'm going to write. But...
Probably, you should rethink your Ultra. As it is very wide, you don't need such long serifs, which only accentuate the horizontal axis. They're diagonal, so their corners will be visible this way or another. I mean, if you close the inner space, serifs could be shortened as well.

The a and s have similar terminals in the regular weight, but different in the ultra. Refine all ultra glyphs first, check them in text; then interpolate.
It is very easy to find wrong ultra letters in printed text, because they will look like heavy black dots, or grayish crocks.

The best way (in my opinion) to find the bold, is to insert single words and longer sentences into a page set with your regular. One copy for one intermediate. Then you can show these pages to your friends, and probably they won't say "this" or "that", but something like "it should be between this and that". Then you decide.

Thanks for the thoughts Marcin! I think you are correct on all accounts, although this was just a rough test. This bold is really just too bold; no one in their right mind would use it in text, unless they were already using an ultra weight in text and needed a bold emphasis. That said, I think a little bolder than the 5% interpolated test would make a good bold for normal use, but like you said, testing in a block of text would be my best bet for the bold weight.

I think I'll still have the "Way Too Bold" version, but I might close up the counters a but more, and definitely shorten those serifs. Even then it might still be too bold, and I will probably need to make the stems a little thinner.

I bet people will love and use the "Way Too Bold" in texts. A broad range of weights is a plus, not surplus. Such extra can be helpful whenever a designer needs to put a special "wayfinding emphasis" into a complex text; it can be used as a paragraph intent replacement, etc.

Thanks for your comments Marcin! You've encouraged me that more weights can be very useful! I suppose my reasoning is that, except for a few situations, I don't even use bold weights, just semibold. But I think there are certain aspects of using different weights that can be a very big plus. I definitely think you are right about "wayfinding emphasis", I have just never designed anything as complex as that.

For everyone interested, I've searched Typophile for "Interpolation", and managed to find these bits in a thread on interpolation and light weights:

Hrant's interpolation reccomendations:

1) Make a midling weight that looks good for text.

2) Make a Bold "candidate" (maybe using auto-weight-gain + mucho tweaking).

3) Extrapolate the lightest and darkest extremes. 

4) Clean up the extremes real good.

5) Make new intermediates and compare against the originals. 

6) Tweak only the extremes to make any new intermediates come out good, dumping the original intermediates.

How Tim Ahrens interpolates:
1) Prepare an ultralight (maybe even stroke weight 0) monochrome weight. 

2) Extrapolate the extrabold. This will be too "proportional", i.e. the stroke weight is constant, makeig a and e too bold. 

3) Modify and clean up the extrabold. 

4) Re-interpolate the regular wight. 

5) Create an MM font with the "true" regular and the interpolated regular wight. 

6) Extrapolate a "triple true" regular (truer than true!). Now you can see what makes the true regular true and make conscious decisions about that. I think the most important thing with designing is that you do everything consciously. 

7) clean up the triple true and re-interpolate the true regular. 



When I generate the final instances I put all three weights in one MM font. The semibold is not only an interpolation between regular and bold but I add a certain amount of extrapolation (regular and light).

The reason for this is that a point does not move on a straight line on its way from light to bold (that is why we need a true regular). However, it should not move on a line with a kink at the regular weight. It should move on a curve instead. So the semibold will not be on the direct line between regular and bold. 
I usually prepare an Excel diagram that shows the amounts of light, regular and bold in all the instances. Again, the graphs schould not have kinks. The semibold usually has about -10% of the light weight, 60% regular and 50% extrabold.

From James Montalbano:
To make a bold ya gotta move either east or west.

I think James' comment really speaks volumes about bolds in some ways, at least, Frutiger used this method, and it worked wonderfully:

This seems like an incredibly good and simple solution; the problem is that a bold really wants to be a little more condensed than a regular, for two reasons: 1) To make a bold really be "bold", the space inside should probably be closed up more, 2) the bold should be a bit condensed in order to not be a large amount wider than the regular.

But an improvement in speed and "visualization" of the process can be made with this method

1) Creating an ultra weight using this method
2) Creating condensed versions of both the ultra and regular by

• squishing both versions to say, 80% width
• interpolating/extrapolating the squished regular and squished ultra to get proper stem widths
• tweaking if necessary

3) Interpolating with the condensed version to get the proper Ultra width
4) Determining the bold weight by interpolating between the regular ultra
5) doing necessary tweaking

I think Superpolator could also help this out, I'm planning on purchasing that as soon as possible.

You can also use "Autopsy" to look & learn
how different fonts resolve weights
http://www.yanone.de/typedesign/autopsy/

Video Keynote here:
http://letterror.blip.tv/file/1911264/

Enjoy!

Also check out this thread: http://typophile.com/node/31036
"Where do bold weights come from"

I also have this comparison to show how most bolds are slightly condensed:

Note that Baskerville, whose bold is very unnatural and ugly to my eyes, doesn't follow this rule. It's also a little too bold, in my opinion. It should be an ultra.

Ahh Pablo simultaneous post! That is a tremendous tool!

A quick test. Hairline to ultra.

I used 4 masters: Hairline, Regular, Black, & Ultra. I could have omitted the Regular, but that gave me a bit more control to the forms. As it stands, interpolation between Hairline & Black works.

Please note the weights aren't positioned along optimal points in the system - I just did 50% between each master.

Not as good as Frutiger but impressive nonetheless!