Note: what I give here is descriptive, not prescriptive, so the notes are not enough to actually create fonts for these symbols. The standards have prescriptions on size I will not mention here. However, I made an attempt to do some faithful rendition.

Character recognition fonts are divided in two groups, Magnetic Ink Character Recognition (MICR) and Optical Character Recognition (OCR). However, that distinction is artificial, some applications use magnetic (actually, easily magnetizable) ink for OCR, while the MICR fonts can be recognized optically.

The history of these fonts comes basically from the financial world. With automatic processing of checkes at other than the native banks, automatic reading of information became important. Moreover, a barcode would obfuscate what was actually written, so in the fifties the first font (E-13B) was deviced that would result in nearly unambiguous reading. In banking generally magnetic reading is used because that is more reliable.

I describe the following fonts here: two MICR, E-13B and CMC-7 and two OCR, OCR-A and OCR-B. All four are used in banking. E-13B most dominantly in the US and Canada, CMC-7 in France and other countries, the OCR fonts in various other countries.

E-13B

This is the first font used in automated banking. It contains the digits and four special symbols used in banking. The four additional symbols are called (from left toright): "dash", "amount", "on*us" and "transit". The font was specially designed so that magnetic pulses would be read unambiguously. That is the reason for some of the heavy black features of some symbols. For this purpose a grid of 7 by 11 squares was used that either had to be white or black. The filling was done so that a magnetic scan would give a pulse signal that was very distinctive.

CMC-7

This font was designed in France to encode more than simply the digits and some special symbols. It encodes the digits, the letters and five special symbols. It appears well-designed considering the then current technology, although some letters have something strange ('G'). And it can also be seen as a barcode. Each symbol is encoded by seven vertical bars separated by small and large spaces. The digits and special symbols have two large spaces, the letters one or three as in the following table: 

	symbol	pattern
	  S1	WNNNNW
	  S2	NNWNNW
	  S3	NWNNNW
	  S4	NNNWNW
	  S5	NNNNWW
	  0	NNWWNN
	  1	WNNNWN
	  2	NWWNNN
	  3	WNWNNN
	  4	WNNWNN
	  5	NNNWWN
	  6	NNWNWN
	  7	WWNNNN
	  8	NWNNWN
	  9	NWNWNN
	  A	NWNNNN
	  B	WNWNWN
	  C	NNNWWW
	  D	WNNWWN
	  E	NNNWNN
	  F	NNWNWW
	  G	WNNNWW
	  H	WNWWNN
	  I	NNNNNW
	  J	WNWNNW
	  K	NWWNWN
	  L	NWNNWW
	  M	NNWWWN
	  N	NNWNNN
	  O	WNNNNN
	  P	NWNWWN
	  Q	WWWNNN
	  R	NWWWNN
	  S	NWNWNW
	  T	NNNNWN
	  U	WWNWNN
	  V	WWNNNW
	  W	WNNWNW
	  X	WWNNWN
	  Y	NWWNNW
	  Z	NNWWNW

OCR-A

By the end of the sixties the time was ripe for full character recognition. The first font designed for this purpose was OCR-A. Of course some characters had to be distorted to simplify the process. The first standardized character set had the digits and letters, plus a few more letters with diacritical signs for other western languages. In addition there were three special symbols used for banking purposes (nicknamed "hook", "fork" and "chair") two additional currency marks (pound and yen) and a series of other symbols from the ASCII character set. Note that the vertical bar is too long to actually be a representation for the ASCII vertical bar, it had it's own purpose. In 1977 the set was augmented with a series of additional symbols and the lower case letters, but those are not shown here. Also some forms were slightly changed because they gave problems with impact printers (like holes in the paper). The symbols as shown here are "bitmapped", so they are not a true rendition of the font (it can not be done in bitmaps).

OCR-B

Later (during the seventies) it was found that technology was able to decode more complex symbols than before, so a new font was designed (by Adam Frutiger in this case). This is the OCR-B font that can currently be found nearly everywhere. An EAN barcode is accompanied with it's decoding in OCR-B. Interestingly, when it concerns a book the ISBN is encoded in OCR-A. As before, the bitmaps shown here do not full right to the font, and many more symbols are possibly defined. The set is as for OCR-A but with lower case symbols added and many more of the ASCII symbols. In fact, the ASCII set is complete. In addition further symbols have been added (and standardized), mostly accented letters, but also some special symbols.

Currently the technology is available to recognize any font when it is printed in a regular fashion. However, even with the techniques as they are used now, errors are frequent and the scanning has to be "eyeballed" to ascertain whether the scanned result is indeed what was intended. Obviously that is not reasonable in banking situations (in the US there are cheques that are scanned some 20 times for sorting), so in those situations the standard fonts that can be scanned reliably when printed to standards are still used. And also, Frutiger's OCR-B font is in my opinion one of the most beautiful fixed pitch fonts available. I always used it in articles and reports when program code had to be inserted.