golfomat's picture


I started using ClearType on my WinXP Machine to the beginning of this week.

So now I'm really torn up, because the lack of readability (legibility?) this fuzzy thing causes, is definitely compensated by the new variety of forms I recognize on my Monitor.

The first time you can really spot significant differences between typefaces, without having to do some workaround using InDesign.

I think this topic would have been worth a discussion - at least I couldn't observe any while digging through the typophile archives.

I'm sure some of you might have a statement on this. Or is it such a No-Go, which it isn't worth debating about?



John Hudson's picture

I don't think this thread really belongs in the Build section, which perhaos is why no one has responded to it. It doesn't, so far, touch upon the implications of ClearType rendering for making fonts.

There is a 'sweet spot' for ClearType rendering which is somewhere around 130dpi. At this resolution, the pixels are small enough for the visible artifacts of the subpixel colour filtering to largely disappear. Below that resolution, e.g. at the old Windows norm of 96dpi the physical pixels are too large for the technology to produce optimal results: the text will still look more faithful to the individual type design, but the fuzz is likely to be noticeable. I used a Toshiba laptop with a native resolution of 133dpi for the past couple of years, and ClearType works very well on this machine. I recently switched to a Dell with a resolution of 146dpi and, of course, ClearType looks even better. I've also seen it in use at 200dpi, and this is really splendid: actually better than quite a lot of what passes for printing these days.

Note, also, that you can tune the ClearType rendering system on your computer to get the best results for a) your monitor's native gamma and b) your personal colour sensitivity. See for more information. Note that you need to be using Internet Explorer to perform this action.

Si_Daniels's picture

>So now I'm really torn up, because the lack of readability (legibility?) this fuzzy thing causes,

The obvious cause

golfomat's picture

Thanks John and Simon for your answers. I think I'm one step closer now.

Sorry for putting this to the wrong category. It is a rather technical issue and I always understood the "build" area as the typophile-tech-centre.

I'm using a CRT Monitor. Would it make a difference if it would have the suggested 130dpi?



P.S.: A great feature for the forums would be posting by hitting Alt+s, like when sending out emails... I tried this just the second before, but it didn't work. Damn, I have to get away from the screen.

hrant's picture

> 5% of the population that doesn

John Hudson's picture

Dominik, ClearType is really designed as a technology for LCD screens. It works by effectively increasing the x-direction resolution by addressing the individual RGB sub-pixels in an LCD pixel. On a CRT screen, the pixels are not composed of discreet sub-pixels; rather, the colour value of the whole pixel is adjusted to display different colours.

Interestingly, and for reasons that are not completely understood, many people do find a slight improvement in font rendering using ClearType on a CRT screen, as compared to traditional greyscale antialiasing. The effect is quite minor, and the best guess as to why it seems an improvement is that ClearType is leaking small amounts of colour into adjacent CRT pixels, and this makes curves smoother than with traditional antialiasing. Because CRT is not the target technology for ClearType, little study has been made of its affect on CRT screens.

John Hudson's picture

One last thing: about CT versus non-CT comparisons, what has always seemed hoaky to me is the use of italics type to exhibit the advantages of CT. We rarely read italics type continuously.

Using italics is useful to show immediate and obvious benefit of ClearType to the rendering of traditionally difficult diagonal strokes. It is pretty obvious that ClearType makes italics more readable, because it maintains the structure of the letters, which is broken in the stepped b/w bitmap, and the notan, which lost in greyscale.
B/W, greyscale, ClearType
So that is a basic improvement in both the aesthetic and functional aspects of one particular kind of type on screen, and it is an obvious improvement easily relatable to the increased x-direction resolution, which has made it useful in showing how ClearType works.

Personally, I tend not to use italics in my comparisons, because I already know that italics are going to look better and be more readable under ClearType. What is of interest to me is tracking how improvements in the ClearType renderer, increases in screen resolution and, in the case of the CT Font Collection, sensitive design, result in experiential gains in on-screen reading.

hrant's picture

> Using italics is useful to show immediate and obvious benefit of
> ClearType to the rendering of traditionally difficult diagonal strokes.

My point is that putting forth examples of the improvement in italics too prominently equates to misleading people. They make you go "ooh-aah" and implicitly pretend that most text will benefit that much, when it won't.


Si_Daniels's picture

>Simon, that seems extremely optimistic, unless you're using "doesn't like" very differently than me...

This is the Advanced Reading Technology (ART) team's numbers based on their research. I can ask them for references supporting this number.

Perhaps you're including people viewing CT on CRTs, non-native resoltions, substandard CT clones, analog LCD's or BGR displays?

However it should be noted that amongst some groups (eg. 'software developers' or 'graphic designers') the % of dislikers may be a bit higher. ART team is working to better understand this group of users. It

John Hudson's picture

They make you go "ooh-aah"...

Well, when you want to persuade Bill Gates to give your department a lot of money so you can conduct R&D work in screen readability and, eventually, pay a group of type designers to make seven multilingual font families for you, a bit of 'ooh-ahh' can go a long way. :-)

hrant's picture

Here's my first attempt at a grayscale italic glyph, with the "h" from John's 3rd row tacked on at the end:


You lose some smoothness, but the crispness is better, and you avoid color fringing. Also note that I'm using up the same surface area, and the vertical proportions (hence readability) are the same, but my apparent size is larger, and the steeper angle helps differentiation (although it's a bit steep for continuous reading).

But clearly it's hard to get italics to look really good. What I might suggest is sort of parallel to John's desire to dump the legacy 72dpi world: do the same to the lagacy world of italics! :-)


John Hudson's picture

This looks good, Hrant. But you have made it a bit easier for yourself by making the stroke weight of the roman two-pixels, and effectively two pixels for the italic too. The effective weight of the ClearType is somewhere between one and two pixels, which I think is preferable.

You may find the image below interesting. I'm posting it simply for information, not to make any point other than the obvious one: there is no grey, there is only colour. Actually, my original intent in the other CT thread was to simply provide illustrations and let people make up their own minds, but then I realised that the illustrations needed explanations.
Zoom without and with subpixels

hrant's picture

What's interesting is that you can do a tradeoff between the "integerness" of weight, and crispness. In an italics, since you have to give up very crisp forms anyway, you actually acquire the ability to deviate from integer weight wihout any [additional] penalty. So I can make my italics slightly lighter or darker pretty easily. However, you're right to imply that a proper relationship with the roman has to be maintained, and making the italics too light would be bad (especially since my angle* is already strong).

Some related stuff:
- I like to have stroke contrast; and -as I implied above- a 2-pixel stem font isn't necessarily too dark since it can have 1-pixel horizontals;
- I could make the Roman lighter too (if also less crisp);
- Using a lighter weight for emphasis in running text isn't so bad - not as bad as the other way around.
- Crispness can actually be a differentiation technique. It's possible that the best companion to a grayscale Roman is a b&w italic, or... shudder... a b&w upright cursive!

* The angle is actually dictated by the medium to a fair degree, btw.

Then there's the issue of what weight looks good - and that depends on effective resolution greatly. I think the color in your "best" (3rd row) italic is very good. But maybe a 2-pixel stem weight with looser spacing would be comparable? Anyway, ideal color needs to be worked into the equation with everything else, like chromatism, crispness, etc.

> there is no grey

True, but you could say that the three channels in a [properly calibrated] display are defined as producing no color when lighted with equal [numeric] intensity.


John Hudson's picture

The angle is actually dictated by the medium to a fair degree, btw.

Optimum italic angle certainly is. ClearType allows for a greater range of angles to be handled better than full pixels, simply because of the increase resolution. In Constantia, I wanted the optimum anngle, so I have a run of 1:3. The other CT Font Collection families have a variety of italic angles.

you could say that the three channels in a [properly calibrated] display are defined as producing no color when lighted with equal [numeric] intensity.

You could say that, but you would still actually be looking at this:

Really big pixel :-)

vincent_connare's picture

depends it's not always the same, sometimes it would be RGB, BGR, or landscape or portrait. So who knows.

Frank Jonen's picture

I think this little app will help to understand CRT and LCD stuff a bit better, though the author states that it's not 100% accurate.


hrant's picture

> I have a run of 1:3

It looks more like 1:4 to me. Mine is 1:3.
In grayscale bitmaps (again, not the same thing as whole pixels), anything lower than 1:4 crosses into too much fuzziness (in my view).


John Hudson's picture

Oops, yes, you are right. I remember now that I experimented with 1:3 (18 degreees off vertical), but decided it was too much. I settled on 11.7 degrees off vertical, which as you say is closer to 1:4, after testing rendering. One of the nice things about the CT Font Collection project is that we were able to test in the specific conditions of the target rendering system.

Syndicate content Syndicate content