The keys were mounted on top of a wedge-shaped frame, with the machinery inside. At the far (back) edge of the wedge was a carriage that was almost twice as wide as the main "wedge"; it shifted back and forth primarily to accommodate the shifts required for multiplication and division. Dials with numbers on them to show results showed through rectangular cutouts in the carriage cover.

There was a number of keys to do various operations; the + key was always prominent. Pushing a key to do an operation unlatched a single-revolution clutch and started a governor-controlled "universal" motor, which was geared down. The main driveshaft (except in the later Marchants) turned once, and then the clutch disengaged. One cycle would add the number in the keyboard into the corresponding dials of the carriage, wherever it was positioned; carries were done in the later part of the cycle. This set of dials was the accumulator, which showed products, sums, and differences.

The carriage also had a second set of dials, which counted cycles of the machine. "Add" cycles incremented the count, and "subtract" cycles decremented it. As the carriage shifted, different dials in this "cycle counter" aligned themselves with the active part of the machine; only one dial was directly operated at once. Except in the Monroe, there was a full carry/borrow. Quotients (and in a few Friden models, square roots) appeared in these counter dials. Once a multiplication was complete, the multiplier showed in these dials.

While other (primarily European) calculators operated according to these principles, their physical details might have differed.

The calculators I have in mind are referred to as "rotary"; there are others, which I don't know much about. Burroughs and perhaps Victor might be among the principal U.S. manufacturers. The distinctive feature of rotary calculators is that their internal calculating mechanism itself (as distinguished from "control" mechanism) rotated continuously. The others (as with at least some cash registers) had calculating parts that went back and forth; the author knows little about these.

If you held down the + bar on these machines, and set the keyboard not to clear at the end of the first cycle, the number contained in the keyboard would continuously add into the accumulator, and the counter dials would increment by one for each count. In all three makes, the main driveshaft(s) would in such a case turn at a steady speed. In the Friden and Monroe, the mechanism would alternately increment the accumulator by the count in the keyboard while noting carries; this phase was referred to as "digitation" by Friden. This was followed by entering the carries, toward the end of the cycle. (Carries could cause carries, so the carry process was sequential, starting at the least-significant digit, at the right.)

The intermittent movement of the accumulator dials in the Friden and Monroe meant that they had to be locked against overshoot as soon as digitation (and immediate incrementing to add in carries) was complete. This was done by a Geneva-drive-like mechanism (see a book of mechanisms; look for intermittent drives.)

The Marchant was different; holding down the + bar meant that almost every moving part was turning at a constant speed; this required added complexity, but permitted the machine to run about twice as fast as the other two. Carries were done by 10:1 reduction gearing from a given low-order accumulator dial to its leftward neighbor. The potential misalignment was "fixed up" when the accumulator "came up for air" at the end of a cycle and disengaged from the main body of the machine. In effect, there were three main driveshafts in the Marchant.

The Monroe (and, I'm almost certain, the Marchant) driveshafts went in reverse to subtract; I don't recall how the Monroe always did carries after digitation, but it was simple.

The Friden main driveshaft always rotated in the same direction; the reversal for subtraction was a very simple scheme that shifted a pair of facing 10-tooth miter gears along a (square) shaft. These gears straddled a matching 10-tooth miter gear on the bottom of the accumulator dial shaft. (Friden accumulator dials had shafts that "leaned" toward the operator about 30 degrees from the vertical. They had knobs on them.)

Many, if not most, of these calculators did automatic multiplication and division. The Marchants I describe (there were earlier "stop-go" types, I am told) were so fast that there was no need to enter the digits of a multiplier into a mechanical storage device ("memory"); you entered the digits sequentially. Even all-9s was not sluggish.

The Friden, in particular, gave you manual access to some normally-automatic aspects of the machine. You could choose to disable automatic clearing of the accumulator and/or counter dials, or any keyboard column. On models with fancy accumulator dials, you could make the dials reset to non-zero digits when the accumulator cleared. You could turn the accumulator dials with small knobs any time the machine wasn't cycling.

The Friden could add or subtract a product to/from the contents of the accumulator. Some Monroes could do quite a number of useful things automatically.

Handling of decimal points in most of these machines was relatively primitive; in the rotary nonprinting mechanical calculators, to my knowledge, only the last of the Fridens had a decimal point key, and that was only for the multiplier. Friden square root did require careful handling of decimal points, but as in just about every other operation, the operator had to preset mechanically-movable markers to show where the decimal points were. The procedure was hardly that; it was made just as simple as it could be.

Indeed, the decimal-point key that appears on every electronic calculator requires enough internal logic that it was not provided on the first Friden desktop electronic calculator; it would have complicated the design too much! (That machine had a selection of fixed points.)