Tiny3x3a

Share:
by Michaelangel007
Cloned from CHECKER by flow-4.

Download disabled

The designer of this FontStruction has chosen not to make it available for download from this website by choosing an “All Rights Reserved" license.

Please respect their decision and desist from requesting license changes in the comments.

If you would like to use the FontStruction for a specific project, you may be able to contact the designer directly about obtaining a license.

Based on "Checker" cleaned up glyphs to be more readable and added lowercase.
Info: Created on 21st June 2012 . Last edited on 12th July 2015.
License Creative Commons
Categories:
  • -
Sets:
Tags:
Fave Tags:
  • -

32 Comments

Duplicates in the digits : 1, 2, 5. (But you can fix this quite easily by copying the other FontStructs.) Great 1st attempt(s) with the lowercase. Though, honestly, it'll take you a lot (really !) more than a single day to design it accurately and without any mistake, since yours is made of numerous dupes : c=a, e=a, h=b, o=a, s=a, w=m, x=a, z=a. Besides g, j, p, q and y are 3x4(!), which is cheating (with the same issue in three punctuations, and there are dupes there too). Not sure whether Small Fonts (Windows) behaves less faithfully with the minuscules, but your effort was appreciated, and any update is very welcome, since it's the most complete micro font I could pick on FS up to now (2+ dozens of similar 3x3 tries, should make you somewhat proud).
Comment by dpla 8th March 2013
PS. Forgot to mention your uppercase set is almost perfect (from what I experimented myself and could see elsewhere)… :-)
Comment by dpla 8th March 2013
A couple of points about font design (pardon the pun) since you seem relatively new to the field of computer graphics, rendering, and font typography:

While it is true that this is a 3x3 uppercase font it is _impossible_ to design a _Monochrome_ font within this context that guarantees each _lowercase_ character has a _unique_ glyph; there simply isn't enough bandwidth (i.e. pixels AND bits-per-pixel) to make each lower case glyph visibly distinct. e.g. Each lowercase glyph would need to be 4x3 not 2x2.

I agree that labeling the 'baseline' while using the 'leading' line for the lowercase descenders is slightly 'cheating'. :-/ However, this font name conveys the width AND height (in pixels) without the leading / spacing to be consistent.

Other font sizes DO include the leading BUT do NOT include the glyph WIDTH spacing, thus they are using inconsistent measurements.

Even worse, unfortunately, there is ambiguity as there is no standard way, or "units", to accurately show AND know _both_ numbers -- the actually pixel coverage AND the TOTAL width or height including spacing -- sadly too many people just don't care. :-/

The 1, 2, 5 digits are not duplicated; the 2 and 5 have been mirrored.

Did you mean the lowercase L 'l' and one '1' have been duplicated? This is a 'Sans Serif' font; again the problem is lack of bandwidth. :-(

The 'S' and '5' glyphs are the same due to being limited to a monochrome 3x3 bitmap.

I'm not sure if you realize it, but with a 1-bit bitmap it is _impossible_ to make 2x2 unique glyphs for 'a', 'c' 'e', 'o', 's', 'x', 'z', along with distinct 'm' and 'w'. Your dupes comment is unfounded.

This criticism is not directed at you personally but to point out WHY almost every font system is garbage. People _assume_ glyphs are _only_ 1-bit HOWEVER that is an incorrect premature optimization! Now IF FontStruct allowed a _proper_ 8-bit alpha/luminance/intensity then it should be possible. The root problem is that the Windows/OSX/Linux TrueType rasterizer does not accept gray-level hints as input; it only generates them as output; TrueType is fundamentally _broken_ for ultra small fonts.

Color hacks such as "ClearType" and fudging the font alignment are needed when dealing with sub-pixel font rendering to overcome this broken design.

Given the constraints of monochrome glyphs this font was designed to _approximate_ a 3x3 font AND still be have readable uppercase fonts! I knew the lowercase readability was impossible as I had to quantize to the pixel grid. That is not a cop-out, just pointing out that ANY font this tiny is forced to prioritize 'approximation' over 'accuracy'. :-(

Thanks for the feedback though. Glad you liked the uppercase glyphs.
Comment by Michaelangel007 11th March 2013
Addendum: The minimalist font size that guarantees uniqueness for lowercase 1-bit fonts is a 4 pixels height. There are examples of the HP48 text editors that demonstrate people's creativity trying to solve this difficult problem!

* http://www.hpcalc.org/hp48/apps/editors/


Anyone interested in font design will probably also be interested in these links:

* Typography Terms
http://upload.wikimedia.org/wikipedia/commons/3/39/Typography_Line_Terms.svg

* Sub-Pixel Font Rendering - How it Works
http://www.grc.com/ct/ctwhat.htm

* Perceptually Tuned Generation of Grayscale Fonts
http://infoscience.epfl.ch/record/99793/files/ptggf.pdf

* FontFocus Whitepaper
http://artifex.com/FontFocus.pdf
Comment by Michaelangel007 11th March 2013
« new to the field of computer graphics »

30+ years isn't enough ???
Comment by dpla 12th March 2013
« is _impossible_ to design a _Monochrome_ font within this context that guarantees each _lowercase_ character has a _unique_ glyph »

Forgive me if my calculations are invalid :

[*] : 1x1 = 2^0 = 1 unique glyph (blank apart) ;


[**]
[**] : 2x2 = 2^4 = 16 unique glyphs.


[***]
[***]
[***] : 3x3 = 2^9 = 512 unique glyphs.

There's plenty of possibilities as you can see (most of them being just trash as characters).

My unreleased 3x3 is beautiful (so you are going to make me think of releasing it commercially… be careful ;-)), COMPLETE, and except for a few characters that are not legible (people will have to decide once forever, since I'm not a judge), it has the solutions you could not find because you could not work as many days/years as me on it. You must be very new to this world. How many effective weeks of hard work ? Did you study all the Unicode charts like me to resize them ? No ? So you are just drawing pleasant and incomplete fonts, hence nothing usable in the IT.


3x4 = 4096 unique glyphs. By reducing some of my lowercase, still quite legibly, I could get, as already stated elsewhere, hundreds of unique glyphs, i.e. ASCII + Extended ASCII (Most ISO variants) + Cyrillics + Plenty of math signs + A great deal of punctuations + Arrows + tiny exponents etc. etc. Believe me, you never tried seriously (which is a lot of thinking and artwork - on sheets of squared paper, not on FS, it's far more efficient).

BTW. when I remarked elsewhere that 5x5 is huge (a fake 3x3 or 3x3 instead), it was not spiteful ; it's 33'554'432 unique glyphs ! Believe me or not, you cannot be serious. Try again.
Comment by dpla 12th March 2013
« The 1, 2, 5 digits are not duplicated; the 2 and 5 have been mirrored. »

Let's see if I went here to comment for nothing. :/

1===l
2===Z
5===S

What is your explanation now again ?
Comment by dpla 12th March 2013
« Did you mean the lowercase L 'l' and one '1' have been duplicated? This is a 'Sans Serif' font; again the problem is lack of bandwidth. :-( »

My eyes can only see what they can see, so if you fill a grid with exactly the same number of pixels and with exactly the same [x,y] positions, I cannot tell the difference. That's what I called a 'duplicate'. (My word may be wrong, but I borrowed it to the bitmap wording, as micro fonts behave the same, you know : automatic duplicates softwares - absolute clones that's what I meant.)
From a coding viewpoint, computers don't care. You can use the same glyph everywhere, the textfile would remain the same, so you cannot rely on 'puters.
Comment by dpla 12th March 2013
« The 'S' and '5' glyphs are the same due to being limited to a monochrome 3x3 bitmap. »

NO : due to // your // limitation of thinking about the problem. You surrendered, hence my comment of encouragement. You need the solution ? Here it is : it's obviously, you have to try to redo S, 5, or both. And you'll see that there are possible variant shapes. And you'll see that it's like e.g. the 'taquin' logical game : http://en.wikipedia.org/wiki/Jeu_de_taquin
(or better Google Image it, you must know it),
where you'll have to strive till the end.
This said, you'll have to keep a sense of style, unlike heterogeneous samples of micro fonts I could see on FS.
Comment by dpla 12th March 2013
« I'm not sure if you realize it, but with a 1-bit bitmap it is _impossible_ to make 2x2 unique glyphs for 'a', 'c' 'e', 'o', 's', 'x', 'z', along with distinct 'm' and 'w'. Your dupes comment is unfounded. »

'a', 'c', 'e', 'o', and 's' can quite well be turned into 2x2, individually ;
I did so in my unreleased 3x4 complete font (ASCII speaking etc. - see above).
They're quite legible once you know the trick, and logical also. That's why I adopted them.

BTW, you are not a great admirer of your friend's works in their micro fonts experiments (or you just overlooked their last stuff/updates), else you would already have found most of their solutions (many years after me, but who cares). The proof that you are wrong is on FS. It's enough tips for you today.

Of course, if you don't think a lot, you think it's impossible/unfeasible/unfounded.
Comment by dpla 12th March 2013
« […] _impossible_ to make 2x2 unique glyphs for […] 'x', 'z', along with distinct 'm' and 'w' »

Agreed for both. (My last free tip here.)
Comment by dpla 12th March 2013
« This criticism is not directed at you personally but to point out WHY almost every font system is garbage. [… till the end of §] »

Actually, I need (and made) a universal 3x3 font (i.e. complete in ASCII, and as legible as possible - I'm no magician, we'll have to educate the masses ;-)). So 'your' system troubles, though quite interesting, is beyond my work. Mine, is universal. E.g., as others do, you can use any medium as a 'rasterizer', e.g. objects on a tiled floor, 3x3 windows etc. Do you get the difference of my artistical viewpoint. But as a coder, of course, I need this font to be usable. Bitmap fonts (real ones) were adapted. A pity if what you say is an issue. And if we multiply the size of the blocks ?
Comment by dpla 12th March 2013
Just in case of, vector-engine related issues in not the point for me. No matter if your blocks are 1000x1000, if it can help the rasterization. All what we need is a fairly good 3x3 grid of 'dots' (replace the last word by whatever you want, be it virtual or physical.)
Comment by dpla 12th March 2013
« Thanks for the feedback though. »

Thanks for yours.
But I'm not dead.
Comment by dpla 12th March 2013
« ddendum: The minimalist font size that guarantees uniqueness for lowercase 1-bit fonts is a 4 pixels height. There are examples of the HP48 text editors that demonstrate people's creativity trying to solve this difficult problem! »

My gosh ! Do you really think that I ignored it, after years of work and searches ?! you should blush if you knew ! If you look very well on your link, there's only 2 fonts out of 74 that might look interesting for me about their very small size. // BUT // their lowercase are at least 3x3, hence the size of an expected uppercase (because the max height is 3 px). That's a lot TOO BIG ! Ignore this HP page (and 99 % of all what is extreme micro font related BTW), it's useless as hell.
Comment by dpla 12th March 2013
« guarantees uniqueness »

PS. that's totally wrong, as I showed you above. BUT to be legible, 3x4 in the minimum, yes.
Comment by dpla 12th March 2013
Alright, I stand completely corrected as I see you have experience and given this some thought.

You know, the fastest way to motivate a geek is to tell him "You can't do that." ;-)


I have updated the 'a', 'l', 's', 'z' glyphs after writing out all the 16 1-bit permutations for 2x2:

[__] 0000 (reserved ' ')
[__]

[__] 0001
[_X]

[__] 0010 (reserved '.')
[X_]

[__] 0011
[XX]

[_X] 0100 (doesn't touch baseline
[__]

[_X] 0101 (reserved 'i')
[_X]

[_X] 0110
[X_]

[_X] 0111
[XX]

[X_] 1000 (doesn't touch baseline)
[__]

[X_] 1001
[_X]

[X_] 1010 can't use -- can't distinguish from 'i'
[X_]

[X_] 1011
[XX]

[XX] 1100 (doesn't touch baseline)
[__]

[XX] 1101
[_X]

[XX] 1110 'r'
[X_]

[XX] 1111 'o'
[XX]

As you correctly state there are 3 unusable glyphs 0100, 1000, 1100 due to the fact that the bottom of the glyph MUST touch the baseline.

Likewise the 2 glyphs 0001 and 1010 must also be rejected since they cause incorrect kerning and whitespace.

There are 5 already used / reserved ones: 0000=' ', 0010='.', 0101='i', 1111='o', '1110='r'


That leaves a total of 6 potential glyphs for use:

Now, I could see one choosing 0111=a, 0110=s, 1001=z, but that still leaves 'c', 'e' and 'x' with: either 0011, 1011, 1101.

Here is a summary of what we have so far:

0000 ' ' used
0001 N/A whitespace
0010 '.' used
0011 '_' used for '_' ???
0100 N/A baseline
0101 'i' used
0110 's' ???
0111 'a' ???
1000 N/A baseline
1001 'z' ???
1010 N/A whitespace
1011
1100 N/A baseline
1101
1110 'r' used
1111 'o' used

However, which of the 3 remaining glyphs (0011, 1011, 1101) would one use for 'c', 'e' and 'x' that doesn't drastically distort and compromise the geometric outline / silhouette? You would be forcing people to learn a new set of glyphs that are not even closely related to the standard lowercase glyphs. IMHO that is not the right way to go.

With all that said, thanks for pointing out that the 'a', 's', 'z' are possible. It was obvious once I wrote out all the 16 permutations on the graph paper. I'm not sure why I didn't before.

I have also touched up the 'm' and 'w'.

Looks like I must eat humble pie. I'm actually surprised how "good" the new glyphs look. ;-)

Cheers
Comment by Michaelangel007 12th March 2013
Ah, thanks for the clarification 1=l, 2=Z, 5=S.

I've fixed the lowercase 'L'; I don't see how the 2/Z 5/S can be changed -- will need to research/think about this for a bit.
Comment by Michaelangel007 12th March 2013
Cleaned up the C, J, 8
Comment by Michaelangel007 12th March 2013
Cleaned up the 'I' and 'Z'
Comment by Michaelangel007 12th March 2013
Swapped 'c' and 'e' glyphs to have 'e' more readily recognizable based on the frequency usage.
Comment by Michaelangel007 12th March 2013
« Anyone interested in font design will probably also be interested in these links: »

Thank you, very cool.
BUT (sorry), not micro font specific.

Link 1 : Pretty little English reminder.

Link 2 : + more links : http://en.wikipedia.org/wiki/Subpixel_rendering

Link 3 : Was new to me, I mean their release, not the résumé on filters and vectorization. An old paper for old approaches. I'll resume my reading greedily asap., although not quite helpful to micro fonts generation.

Link 4 : I didn't know a lot about GS, and this experiment either (maybe not at all, one cannot remember everyone's studies). The web either BTW… This technique looks smart though. Maybe it's obsolete, or superseded ? (Is the fontFocus patent still pending ? Obviously yes, otherwise they would have updated the Pdf). Because, if they were quite objective, they would have reused all their comparisons under the last chapter : 'Comparison to RGB subpixeling'. I'd be glad ClearType (to cite the most famous subpixels trick) renders as badly as they say. But if you read on Wikipedia, this technology rules the market for a long time, and things are not going to change, thanks to its assets.

I quote the Pdf from Artiflex : (that's not illegal)
«
Advantages of FontFocus over RGB subpixel rendering:
1. • No color fringing.
2. • Need not be tuned for a specific display.
3. • Works on colored text and backgrounds as well as black and white.
4. • Works equally well whenrotated 90 degrees.
5. • Effective on both LCD and CRT technologies.
6. • Works even when resolution is resampled.
7. • Suitable for Web graphics and other cases when client display is unknown.
»
My comments now.

1. Things are going to be always better for ClearType, but as a humble bitmap specialist, I wonder how they hope to get a superior result in FontFocus via 256 grays, whereas ClearType has thousands ??? Don't they know - or just do they hide - that ClearType is greatly useful to font weights, not just their contrast-specific weaknesses. Since they offered just one example of font weight in this chapter, understand that nobody can tell which from the two is closer from the right weight (as printed), which could be a misleading advertising if my suspicion turned out to be founded.

2. Of course with grays (see above). CT's tuning step is not of such an annoyance, and optional (the default parameters are averaged).

3. CT too, at least today. I've just checked in MS Word @ 100 % size with all mentioned combinations.

4. ? (I could take 5 minutes to check this via the graphics card panel, but I have plenty of windows that I need to keep the size and layout. I have already read this - now fixed ? - point elsewhere.)

5. Of course, CT is LCD specific. But hey ! CRTs are not made anymore. I must quote Wikipedia : « Displays that have no fixed pixel positions, such as shadow mask CRT displays, may be harder to read if ClearType is enabled. However, on CRT displays with a similar pixel arrangement as flat-panel displays, such as aperture grille and Trinitron CRT displays, it can result in a slightly improved readability. »

6. ? (I Need more detail about this step.)

7. That's 2+4+5 at least, hence a repetition, not another argument.

My conclusion : quite dubious, hence not a serious reading from an implementation standpoint. Perhaps more helpful in its first chapters, for the theory.
Comment by dpla 12th March 2013
« a geek »

Please, not you too ! :-))
No, I just can't stand the word “impossible” in coding, when all (really all) the possibilities have not been discarded.
Comment by dpla 12th March 2013
« Looks like I must eat humble pie. I'm actually surprised how "good" the new glyphs look. ;-) »

Damn ! If you knew all the sacrifice I did to get a real 3x3 and 3x4 font, rather legible with a // little // practice, you'd die, in return. (Again, I can be very proud of my own unreleased micro fonts ; and so will you after a lot of effort, the main reason of my posts.)

I'll comment on the rest, but you know, the more detail, the more unveiling. Hence my possibility to sell these stuffs (I forgot to be rich in the long process of creation, blame the artists/experimenters). (Oh, I must repeat here that I came on FS to compare my work, which does not mean I'll release it for free if nobody on earth already found something as valuable, with his unavoidable concessions on the legibility, you'll see in your quest :))
Comment by dpla 12th March 2013
« I don't see how the 2/Z 5/S can be changed -- will need to research/think about this for a bit. »

As I mentioned, fortunately, there are already pretty cool variants on FS. I cannot be more precise about the digits, since it's the only secrets, with lowercase, that prevent you/everyone to get a complete and uniform set of ASCII. You'll have the solution once you have downloaded all the micro stuff I already listed on FS. That's a great present, for an advice, you know.
Comment by dpla 12th March 2013
My methodology.

• Size check : 3x5. Now it's not a valid candidate at all. Try a fix, at least to get a 3x4. Did you left the 3x3 scene ??? Don't surrender.

• Visual duplicates check : Of a low interest, when a font is of undefined size… but as a piece of info : c===o, x===c/a, 1===I, 5===S. The 3x3 font is not extremely difficult to get, but as something legible, it is. You are not far from a complete and true 3x3. But I have adopted another approach, which is very interesting for me to compare with yours, since I'm still wondering if I made the right choices. (You'll see that once your 3x3 will be done technically [a prerequisite that I'm waiting for you to post - don't be ashamed about the temp trash, it must have no dupes], you won't be totally satisfied about its legibility, which I think is normal, until I see better choices than mine. If mine are the most valuable, and only after an public approval, my 3x3 would be a candidate to a universal 3x3.) I don't know about your mind. It may sound unfeasible. If so, please follow my prerequisite, it's logical after all.

• Lowcase style/height check : great job ! Individually, there are a few perfections (based on my 3x4 font). This said, you don't have the '80s culture in your veins' (it's not depreciative, you will learn [and this stupid sentence is not vain, which I admit looks the contrary]) ; if you were on these machines, you'd've known that there are conventions on a few letters that a lot of people (dozens of hundreds perhaps) were more or less used to. This advice given, 5 lowercase are out of the 3x3 limit. Is it the main reason why you adopted the 3x4 matrix even in your punctuations ?
Comment by dpla 12th March 2013
Last thing to help you greatly : why such a 2x2 limit to your lowercase ;-)
Comment by dpla 12th March 2013
* Fixed pipe '|' as it was one pixel too high.
* Tweaked '1' so it has a top-left 'pseudo-serif' / "ear"
Comment by Michaelangel007 12th March 2013
> This said, you don't have the '80s culture in your veins'

Considering I was designing fonts on the Apple ][ with half-pixel support in the 80's I respectfully beg to differ. ;-)

My Program7x8 font is heavily influenced by the clean pixel fonts of the 80s.
Comment by Michaelangel007 13th March 2013
Sorry for not replying before ! Thank you for the fix, good job on both chrs. So I bet you stick with the easier/more legible 3x4 grid ? ('|' being 3x4 now. I'll change of folder in my archive soon :-/)

---

Let's talk oldie now.

http://en.wikipedia.org/wiki/Apple_II_series
Apple II series (Apple ][)

Quote : “Considering I was designing fonts on the Apple ][ with half-pixel support in the 80's I respectfully beg to differ. ;-)”
It's surely sincere, but I would not swear it until I see myself you could design fairly //small// fonts on this machine too… Actually, that was not intended to be irreverent at all. Merely a straightforward way to get more clues from yours. On such limited displays, I wondered : how can one have missed or forgotten the tricks that doubled the width of any good word processor ? (the number of columns, that is, of course.) On the computers I owned or tried (rarely Apple in the 1980s), there were routines in assembler that even did the job while we programmed in Basic. This may be trivial, but a lot of people of this era (the pixels-deprived mates) must have seen, even practised the creation of monospace, 3-px width fonts by the way.

What do you mean by “designing fonts with half-pixel support” ? Is it (just) sub-pixel applied to big pixels ?

Did you rename 'Program7x8' lately ? Anyway, I had a quick look at your 'Program5x7a', 'Program5x7b' and 'Program5x7p'. The lowercase letters are almost the same as the ones that were produced on the other 8-bit machines (in a 8x8 grid, that is, there's no miracle), though some of them are still too large to help you decide how to scale them down with confidence, IMO ; precisely the ones that differ in your font from my memories (and from what I chose, temporarily).

But believe me, I did not logged in FontStruct to see the same choices I retained in my forthcoming 3x3 font. Your small letters, yet with descenders that make them 3x4, are a valuable addition to my study. That is to say, I need to know how skilled type designers - like you - see our common characters when they are shrunk to almost the max ; and which parts you remove, even which tricks or substitutes you prefer. As such, I'll need your opinion in the future, when you'll be invited to critic and comment upon my own work. Let me know if you think I'm unfair again (, ol'pal '-})
Comment by dpla 9th April 2013
I gave you special thanks here. :-)

https://github.com/Michaelangel007/nanofont3x4

Thanks again for all the constructive criticism and feedback!
Comment by Michaelangel007 12th July 2015

I did not notice until Google directed me to your GitHub page today…

SORRY for the delay of 2 years! (I still work on this project actively.)

THANKS for the kind line! (We share the same interest and concerns.)

Several duplicated glyphs are still there. (It is quite feasible to get rid of them.)

Your 'font' (pseudo basic ASCII) is on my very long list of reviews/comparisons to be done and published sooner or later. (I can still be contacted in the meantime, but I cannot unveil my thousands of tiny fonts FTM.)

CYA!

Comment by dpla 1st August 2017

Also of Interest

More from the Gallery

MacBethCalligraphyby Michaelangel007
Program5x7bby Michaelangel007
Tiny3x3aby Michaelangel007
Program5x7aby Michaelangel007
cross stitchby tlmccord
Rough Scriptby funk_king
Optill 2Aby Frodo7
Maniacby D-Sheep

From the Blog

News

New Bricks: Square Connectors

News

The Video Game Font Preservation Society

News

FontStruct goes open source!

News

New Bricks: Half Arcs