Comptometer

Invented in the late 19th century, this calculator became very popular and went through numerous versions and improvements. Built as a business machine, the manufacturer, Felt and Tarrant, realized that they could sell more units if businesses could also hire people expert in their use. They opened training schools and taught students (primarily young women) over months to operate the machines by touch at high speed and with near perfect accuracy.

Once trained, the girls entered the workforce at very respectable wages as skilled office personnel. Large companies might employ dozens of girls. They'd occupy a matrix of desks in a large room, each desk built with a recess to hold a Comptometer. And, barring job hopping, new model upgrades, and irreparable breakdowns, the girls and Comptometers stayed paired for the duration of their working life. If a girl stayed with a company until retirement it was common to gift her unit to her. Comptometers were made and used into the '60s.

High speed was made possible because the act of pressing a key actually added the key's value into the accumulator, and because multiple keys (in separate columns) could be pressed at the same time. A proficient operator, or "computer" as they were known, could use two hands to enter/add an 8 (or maybe 10?) digit number in a single operation. I can easily imagine achieving speeds of 2 or 3 adds per second, maybe more. I've read autobiographical accounts of multitasking computers who were able to work and casually converse with the other office girls at the same time.

A simply fantastic site devoted to Comptometers is Biography of a Machine. There is also John Wolff's Web Museum which, has a big and expert section on Comptometers. Both are well worth a visit.

This unit is a Model C. I've opened it up, cleaned and lubricated everything, and straightened out some bent keystems. The calculator works fine for all operations except clearing the accumulator: it clears to 989000000, so to take it the rest of the way you have to press the leftmost two 1 keys.

The reason for this is interesting. The keys press down on long "segment levers" that are hinged at the rear of the box and attached to strong springs that pull them back up. There's a segment lever per digit of the accumulator. The front end of each lever is a toothed rack that engages an accumulator dial. The segment levers turn the dials not on the downpress, but rather on the upswing.

The clearing process involves disengaging the dials and spinning them to 888888889. A bar then descends near the back end of the levers (in back of the 9 keys) to press them down the distance equivalent to the amount produced by the throw of a 1 key. The bar is then raised and the levers spring up adding 111111111 and causing a carry cascade to overflow the accumulator to 000000000.

What's wrong is that the lever below the leftmost key column is very slightly bent. It was bent worse, to the point that the keys in that column misadded, but I was able to remove it and almost completely straighten it. All the keys now press the lever down correctly. But the ability of the bar near the fulcrum of the lever to press down the exact amount needed to move the rack one tooth depends on very precise tolerances. The amount by which the bar presses is magnified by the length of the lever, so the throw there has to be quite small. And, due to the slight bend in the lever, it's not large enough to press the lever quite as much as the throw of the 1 key.

As a result, when the bar lifts it adds 1?1111111, where "?" denotes a value that is neither 0 nor 1. It doesn't cause the 2nd most significant digit of the accumulator to rotate to fully to 9 and as the accumulator dials can't make partial rotations it stays at 8. Worse, the imperfect add interferes with carrying. It jams the carry into the the 2nd most significant digit of the accumulator, which also prevents the 3rd most significant digit from rolling past 9 to 0. As a result, the display comes to rest at 989000000.

Straightening the lever perfectly is probably not doable without specialized equipment. Getting a new lever is out of the question, and pulling one out of another machine would depend on finding a Model C being sold for parts - but I've never seen another Model C for sale. The commonly found J models had a differently shaped lever due to the introduction of extra features to prevent partial lever presses.

My only other option is to shim the bar or lever at the point of contact just enough to get a little bigger throw. That seems very doable, but as the machine is 99% operational and this glitch is so easy to work around I have not felt any urgency to try it.

Comptometer Manual, (c) 1916, 52 Pages