More Adventures in Typography
The last time I told of my experience with typefaces, we learned quite a lot, including numerals and currency signs having the same spacing amount and how to find fonts on Windows and open them up on FontForge to see how things are done. In the many years since, I have learned much, much more, but I've also come upon some more obstacles—all of which I've gained while developing my typeface inspired by Times New Roman (TNR), originally called Vexle, now referred to as Sinna, which I officially started working on in late 2021.
For Sinna, I initially created letterforms based the roman weights of TNR. After the lengthy process of measuring the weights of the original letters, creating the new ones, and saving each individual character as an SVG file, I opened a new FontForge file for Sinna and opened the font file for Times New Roman, to get references for spacing and kerning.
Some cases were very easy, such as O and o, where the side bearings could be easily copied to the corresponding letters in Sinna. The same applied for other characters, such as the parentheses, the diacritical marks, and @, the at sign. The other characters, however, proofed to be a little more challenging. Since Times New Roman is a serif font, many of the characters have serifs that extend beyond the main stroke. Some of these serifs are also at an angle.
To solve the mystery of the side-bearings, I would simply right-click the point, select “Get Info,” and find the coordinates, referencing the x-coordinate (the number on the left section). For such points on the right side of the letter, I would subtract the amount of the whole width from the coordinate.
I discovered the side-bearings of the angled letters, like V, by going into the background layer and drew angled lines to see where they would land if there were no serifs. However, this would cause edits to appear on the FontForge file (causing an asterisk to appear in the file's name on the window), so some advice if you want to try this at home: when closing the altered FontForge file on a font on your system, be sure to click “Don't Save” when the Save Changes window appears. Better safe than sorry.
Finding the side-bearings without the serifs led me to another discovery. When I made typefaces in the past, for letters like H or I, I would have the side-bearings be the width of the main stroke, like 85 em units, but when creating Sinna, I discovered that this wasn't true. When I checked the widths on TNR, the main stroke of H, for example, came to 192 em units. The left side-bearing (discounting the serifs) was 229 em units while the right's was 236 em units.
When researching the italics for TNR for Sinna's italic form, I found out the former's italic FontForge file has slanted lines on either side of the character. Whenever I made italics or obliques, no such lines appeared. Furthermore, my italic characters strayed slightly out of width limits and italics for installed fonts didn't seem to do that. How was this possible? How could I achieve it?
I found the answer by hitting Ctrl+F in the TNR italic file to get the font info. There, I discovered a section called “Italic angle,” where one puts the italic angle for the font. Filling this in caused the slanted lines to appear when I italicized Sinna.
At first for Sinna, I started off with the roman weight, figuring I could easily increase that into bolder forms as well as decreasing it into lighter weights. While the former seemed possible, as I have created fonts with bolder weights before, the latter ended up being a bit more difficult.
I used the Change Weight tool and set the “embolden by” setting to certain em unit amounts less than the main stroke width. This ended up not only reducing most of the width of the characters while leaving some strokes at their original weight, but also moved the top and bottom further away from the baseline and cap height. It kept on doing this no matter how I set up the Change Weight window.
I was at a loss at what to do. In the end, I decided to recreate all the characters not in their roman weight, but in the thinnest weight present on the original TNR characters, saving each individual one as an SVG to import into FontForge.
A new file was created for Sinna Light, then erroneously called Sinna Thin. After placing the letters into the slots, I noticed that the widths between characters seemed to be rather inconsistent, although I made the new strokes the same width in Illustrator.
This was especially strange when looking at letters close up, when they contradictorily appeared to have the same width at a distance while zooming up on them showed their varying sizes.
Another notable weight issue arrived in the form of the form used for commas, certain cedillas, punctuation quotes, and the apostrophe. At first, I redesigned it like a really thin version of the TNR comma. But upon placing it into FontForge, it looked out-of-place, having a notable thick/thin contrast, especially when compared to other characters.
Following a period of not knowing how to solve the problem, I ended up creating a new Illustrator file from scratch for character and width reference during the times at work when there was hardly anything to do. I carefully measured the skinniest width of the letters so that the weight for each of them would appear consistent.
During this process, I wondered where I should place the skinny width reference on the original TNR characters. Do I place them closer to the counters or towards the opposite edge of the stroke? I used some letters from Gill Sans and found out that it was the latter; pretty obvious in retrospect, since for O, I placed the weight references at the top and bottom so that it would touch the stroke and I would be logical to place them near the edges away from the counters.
With the creation of a new comma form, the weight issue was solved. It looked more like it fit in now. The stroke inconsistencies with the other characters disappeared, too. I guess my second attempt at creating the light characters was more precise.
I found out that the kerning for small caps letters is the same as their capital counterparts. For example, in Sinna, the kerning distance between T and O is -37, as it is for T & the O small cap and the T small cap and O small cap. Ditto for other letters like R and T (which is -123) and R and the T small cap, etc., etc.
For accented characters and combined forms (like ij), I uncovered another hack. When creating such characters, I would unlock them after doing so. Consequently, when validating the font or adding extra points, I wouldn't just have to fix the original letter(s) or accent mark, but the character with the diacritical mark as well, adding extra time to the process.
While designing Sinna, for whatever reason, I didn't unlock these characters after making them. When mending and simplifying a certain character, if it was a part of a combined/accented form, the character with the combo or accent would be mended afterward, shaving off some work time. No other action was necessary!
When comparing TNR with Sinna, I noticed that accent marks for the bottom of characters, such as Ḥ and ṃ, were placed well into the negatives, with the anchor point being at zero. Previously, I would place such anchor points at the top point of the character.
For letters like Ľ or ť, I placed the anchor point on the comma mark. I later discovered the slot for the comma-like cedilla and pasted the shape in there. Wishing to feel like a professional typeface designer and developing more ambitious thoughts, I created the horn accent character for letters used in Vietnamese, like Ơ and ư, I modified the comma shape to be similar to the one in TNR. For the hook above character, as seen in ả, I ended up just shrinking the top part of the question mark.
Finding many hacks and solutions to problems doesn't mean new problems didn't arise, however.
After placing all the characters in the Sinna Light font, I figured it would be time to create the next weight up: Sinna Thin. For the light font, I kept the side-bearings from the original roman file. I found a link outside the FontForge tutorial site that mentioned techniques how to embolden fonts on the software. One such way was to have the counters “squish” so it wouldn't deform the shape to give the counters space.
Another method I tried was to have the counters set to “remain,” with Steersman Bold. After doing this, the side-bearings for the characters remained the same as the roman font, so that potential issues with letter-spacing wouldn't occur, regardless of weight. All the straight angles remained straight.
When I did something similar for Sinna, the side-bearing amounts remained, but many of the letters were distorted, notably M, the right stroke of which was slightly diagonal, and T, which had the stem unaffected by the weight increase and notably off-center. Some letters also had the tops and bottoms expand past the cap height and baseline.
I created a new version of the Light style for Sinna, this time expanding the weight while setting the counters to “squish.” The letters emboldened properly this time, with nothing getting off-centered, distorted, or going past the cap height or baseline. On the down side, this caused the side-bearings to be reduced from the thin weight.
When I checked the spacing between letters, everything seemed to be okay, but I nevertheless worried about bolder forms clashing into one another. Maybe the solution here is to change the side-bearings to their previous amounts after altering the weights of the new font.
After trying to embolden Sinna Light Italic to Sinna Thin Italic, I noticed something odd with the accent marks, particularly the circular ones.
Other distortions in the diacritical marks occurred, too. I didn't know how to solve them. I emboldened them when I emboldened the other characters, using the “Squish” setting. I undid the alterations to those characters, at a loss at what to do. I selected all the characters to alter the weights using this method and no such issues happened.
I decided to click on each slot and embolden them individually. I also set the bold setting to “retain” instead of “squish.” It seemed to work. I think I might have kept it at “squish” and just edited it with the character window open. Unfortunately, I don't remember exactly what happened; a side effect of starting up (or restarting) typing this blog entry some time after actually doing the actions on FontForge.
Similarly, when creating the initial italic forms, some of the accented characters had their diacritical marks off-center. Others remained centered. I created the accented characters (with anchor points) before italicizing them. I never left the accented characters blank prior to changing their style before. I'm not sure if I have to alter the coordinates of the anchor points after doing so, or at least before creating accented characters. TNR Italic has different coordinates for its anchor points, so that might be true.
More often than not, when increasing the weights on my fonts, some counters would be distorted. A notable case in Sinna is when I first created the regular font, the counter in Q, would sink a little bit (but still noticeable). It wasn't a major deal; I could just easily move the counter into its proper place.
For Torinus Bold, the thin strokes got a little thicker and I just moved the appropriate points into their correct spot. I think I might have recreated the accented characters after doing so, since I didn't keep them locked. Presently, I haven't found a solution to this yet.
On a final note for weights, when I turned Sinna Thin into Sinna Regular, using the “squish” method for add em units, this had the negative result of mutating some features on the typeface, like flattening areas with hooks and leaving little to no space between strokes. I had worked on Sinna for almost a year—since late 2021 and, after progressing to this point in early July 2022, this caused me to take a sabbatical from the font, at least until I can find out what's wrong and how to solve it.
Perhaps I should take some advice from Ben Whitmore's link above start with extremes, like one file for thin, another for regular, another for bold, etc. I'm unable to take all his advice at the moment as I'm not entirely sure how interpolating fonts works.
Long ago, in an early version of what would be Sinna, I felt ambitious and tried to add Greek and Cyrillic characters, even though I don't write or speak either. Being based on TNR, I copied the side-bearings from the latter font, using the method I've been utilizing of finding the distance of a point and either using it or subtracting it from the right side-bearing.
The tricky part came to the kerning of these letters. Since I'm not a fluent speaker or writer in neither Cyrillic languages nor Greek, I didn't know that there was a lot of kerning for many letters that I needed to do, particularly for Cyrillic characters. It seemed too daunting for me and abandoned the idea of using Greek and Cyrillic lettering in this and future fonts.
Lastly, here's something interesting I learned, but I'm not totally sure is right. I found some figures that seem to affect the superscripts and subscripts. I haven't found anything online about them. But for completion's sake, I replicated the figures into the appropriate section in Sinna Light.
Ultimately, I've learned lots of more information about fonts and typography over these past years, but there are still many mysteries to solve. I learned the correct way to make skinnier weights by placing the weight reference away from the counters rather than towards them. You should also leave accented characters locked after creating them, so that when you validate the font, those characters will be mended too, and it'll save you lots of time.
But I'm still not sure about some cases with making typefaces bolder and issues with italics, especially with accents. Hopefully more knowledge will be discovered in the future; maybe through experience; maybe through internet research...maybe through internet comments (wink, wink, nudge, nudge).