Why is the font used on checks meant for machines to process so odd? It doesn’t look like much of anything, actually, with its oddly bolded segments. I know it was designed for mechanical `eyes’, but did it need to look so bizarre for a specific technical reason?
Couldn’t the machines process a standard font?
MICR characters were designed to be read by magnetic sensors. Each character was designed to have a distictly different quantity of magnetic ink in them, so the sensor would respond to the amount of magnetism and recognize a different character.
More information here.
This brings to mind an old mystery in the banking biz - How do you pronounce it? There’s four ways, and each has its rabid adherents. Wich are you?
mick-er
mike-er
meek-er
em eye cee ar
However you care to say it, MICR is a fairly stone-age technology that survives fifty years later because it works. Bank of America started developing it in the early 1950’s. The characters are so goofy looking because the MICR readers read the symbols with a magnetic head, similar to a tape deck. Only recently, are the characters being read by optical recognition.
Back in the 50’s, BofA was thrilled to be able to read checks at any speed - it beat the pants off having clerks picking up checks and tossing them into piles. Now, check processing machines can scan, read, photograph (microfilm and/or digital imaging) and sort checks at up to around 2,000 per minute.
Rather than reading a literal digit, the head picks up a series of magnetic pulses. The timing of these pulses is translated into digits. Because the original technology was so limited, the MICR character set is limited to the digits 0-9, plus four special characters known as dash, on-us, transit and amount.
More info on the technicalities of MICR can be found here.
You can see true OCR fonts on your own machine. Assuming you’re running windows, run ‘charmap’, or however you wanna check out your fonts. Odds are you have the OCR-A font there. (OCR-B is slightly smoother - there are loads more for special uses, bar codes, etc.)
The reasons for OCR-A are easy to see looking at it, it’s because text recognition is hard enough as it is; You want to be able to recognize a character despite the fact that it’s not going to be absolutely centered, level, or clean. This site has some samples of various OCR fonts.
It isn’t OCR though. It’s an entirely different process called Magnetic Ink Character Recognition (MICR). It’s much faster than OCR, and more accurate, which is why it’s used in banking. The characters are designed to be readable to people as well as the magnetic sensors. The reason it is faster is that the magnetic sensor detects a unique voltage level for each character at the outset, rather than having to go through many hundreds of processor cycles, like OCR does.
And the early MICR machines were entirely mechanical, using no microprocessors at all. If you could figure out which levels made it do what, you could make them do interesting things. There’s a brief discussion of this in the book Catch Me if You Can, [sub]Now a major motion picture![/sub]
I see a lot of MICR where I work. Basically, the machine reads each character as a series of up and down spikes, and the number, timing, and amplitude of the peaks tells the machine which character it’s reading.
As a check moves past the magnetic head, the head sees a changing voltage. The head is perpendicular to the direction of motion, so all that matters is the amount of ink in a vertical line intersecting each part of the character. This means that with some Photoshop work, you can make your own MICR-like characters that don’t resemble digits, but appear exactly the same to the mag head.
When the amount of ink in front of the head increases, the voltage spikes up; when it decreases, the voltage spikes down; and when the amount is constant, the voltage is flat. It’s easy to recognize characters if you watch the mag head’s output on an oscilloscope.
For example, the MICR 0 is essentially a thin square. There’s a sharp peak at the left vertical bar, then an immediate drop and a low voltage in the middle of the character, then another sharp rise and fall at the right vertical bar. The MICR 8 looks a little different, because there are smaller vertical bars next to the main “legs”; the head (reading left to right) sees a small rise, another rise, a drop, a sharp rise, a small drop, and another drop.
Notice that MICR 2 and 5 appear almost exactly the same to the mag head: a small rise, a drop, a low plateau, another small rise, and another drop. The only difference is that 5 is a little wider, so the plateau is longer; if the check reader’s motor speed is a little bit off, it can confuse the two digits.