hardware Internet Latest news Technology Reviews

How Hard Could It Be to Get Millions of Phone Bills Right?

It may be a foreign concept to anyone who has never paid a dime for a phone call over and above the monthly service charge, but phone calls were once very, VERY expensive — especially long-distance calls, which the phone company ungenerously defined as anything more than a few towns away. Woe betide the 70s teen trying to talk to out-of-town friends or carry on a romance with anyone but the guy or girl next door when that monthly phone bill came around; did anyone else try to intercept it from the mailbox before the parents could see it?

While it seems somewhat quaint now, being charged for phone calls was not only a big deal to the customers, but to the phone company itself. The Bell System, which would quickly become a multi-billion dollar enterprise, was built on the ability to accurately meter the use of their service and charge customers accordingly. Like any engineered system, it grew and changed over time, and it had to adapt to the technologies and economic forces at the time.

One of the most interesting phases of its development was the development of Automatic Message Accounting (AMA), which in a very real way paved the way for the wide-open, worldwide, too-cheap-to-meter phone service we enjoy today.

An Intensively Manual Process

From its very beginnings, telephone networks were money-making ventures, or at least were supposed to be. To that end, accounting systems needed to be developed alongside the network infrastructure, and needed to scale with them as the networks grew. That was a pretty simple proposition at first, when networks were generally limited to a single city and where connections between caller and recipient were generally managed manually by switchboard operators. Having humans in the switching loop was really the key to billing, since operators could easily write down information about the call on a ticket for submission to the accounting department, which would determine the charges and bill the customer.

As the phone network grew in scope, so to did its complexity increase. Trunk lines were installed between local exchanges, allowing subscribers to “reach out and touch someone” in the next town over. Automated switches, at first step-by-step (SxS) switches and later crossbar switches, automated away the need for operators to complete local calls — subscribers could now complete calls simply by dialing the correct number. This represented huge cost savings for the phone companies, who no longer had to employ vast numbers of operators, but it also presented a challenge: with no operators in the loop, who would take care of recording call details for billing information?

At first, the answer was simple — keeping some operators on the job. This was pretty much the answer to the problem of long-distance calling up until about the 1940s — subscribers could directly dial numbers within their local exchanges, and even within the small group of surrounding exchanges. Outside of those areas, specialist long-distance operators would make the connection and record call information for billing purposes. It worked, but it was expensive, error-prone, and would obviously not scale well as the network grew, so an alternative had to be found.

The first steps toward Automated Message Accounting (AMA), as the systems developed by Bell Labs would eventually be known, were taken in the 1940s with automatic ticketing. Instead of operators manually recording call information on a ticket, automatic equipment that could create billing tickets began being installed in the central offices of telephone exchanges. These devices sat before the exchange’s outgoing trunk lines and were essentiallly electromechanical printers driven by the impulses created by dialing a number. The ticket printed the caller’s numbers, the called number, the date and time of the call, the overall length of the call, along with technical and subscriber information. The call information was printed in plain text, and tickets were collected from the printers and forwarded to the billing department regularly.

Punch It Up

A section of perforated AMA tape. It’s not easy to appreciate the dimpled shape of the perforations here, but they were critical for proper reading of the tapes in centralized accounting centers. Source: Bell Laboratories Record, vol. 29, pg. 402 (1951).

Automatic ticketing was a huge leap forward for the phone company, but it still had obvious problems — it only partially automated the billing process. Someone still had to sort and process all those billing tickets manually, so while automatic ticketing reduced the need for some operators, it meant more employees were needed in the billing department. A proper Automatic Message Accounting system would require recording call information in a way that it could be machine readable, so that both ends of the billing process could be automated.

The Bell System began deploying a fully automated AMA system in the late 1940s and early 1950s. Rollout of AMA was gradual; if it was nothing else, the Bell company was extremely conservative, and technologies cooked up at Bell Labs were given thorough testing under real-world conditions before being widely deployed across the network. And so it was mainly exchanges in larger cities that got first dibs on the new AMA equipment.

The Bell Labs researchers made some interesting engineering choices when designing the AMA equipment. Given the era, the obvious choice of recording medium would have been the venerable punch card. But Bell engineers decided to roll their own system — literally.

AMA recorders used oil-impregnated paper tape in long rolls as their recording medium. The tape was about 3″ (76 mm) wide, with information recorded as perforations across the width of the tape. There was room for 28 perforations, which allowed for six different fields to be recorded. Each row started on the left with a three-hole entry index field; this was followed by five groups of five perforations, each encoding a single digit from 0 to 9. Each digit was encoded with a 2-of-5 code, where only two “bits” out of the group of five are set for each digit.

Two-of-Five encoding truth table. Each “bit” is weighted as either 0, 1, 2, 3, or 6, and the two positions that add up to the digit’s value are set, or punched in this case. Zero is handled as an exception so that two positions out of the block of five are always set.
The AMA perforator mechanism. Getting the 28 punches to line up 0.1″ apart with such large electromagnets driving them was no mean feat. Source: Bell Laboratories Record, vol. 29, pg. 505 (1951).

The AMA perforator itself was an ingenious design, and one that took great advantage of Bell’s long experience with electromagnetic systems. Each perforator had to cram 28 electromagnetically operated pins spaced only 0.1″ (2.54 mm) into less than 3″ of space.

This was accomplished by arranging the perforating magnets into four groups of seven, each arranged semicircularly on two different levels. Each pin lined up with a cylindrical punching drum, a precision-machined part that sat under the paper tape and allowed the pin to punch a cone-shaped dimple in it. The shape of the perforation was critical because it formed part of the drive mechanism for reading the completed tape; the dimples would engage with a similar drum in the AMA reader in accounting, which would rotate and pull the tape through the machine. This meant that the completed tapes had to be treated gingerly so as to not collapse the dimples before the data could be read.

Since each row could only encode five digits, the complete record of a phone call would cover multiple rows — anywhere from four to six. This presented a problem, since the time from starting a toll call to resolving it — either because the called number never picked up, or when the conversation ended — was variable and often quite long. Even a moderately busy central office could expect a different toll call to be started before the previous call had been completed, which meant that records from one call would often be spread across a wide amount of tape. It was even possible for records of a call to be recorded on two separate rolls, which were changed at around 3:00 AM every day. The AMA equipment in the accounting center had to handle this eventuality, which it did through devices called assemblers that correlated records through a two-digit entry index and the timestamp for the entry. The assembler would then record the sorted records on a separate roll of tape through another perforator, for further processing on the other machines in the accounting center.

AMA Across America

For all its computational crudity, AMA was wildly successful. As the short promotional film below shows, AMA enabled direct-dialing of long-distance calls for the first time, albeit in a limited number of markets, and introduced subscribers to the concept of an “area code.” The introduction of AMA was critical to making the coast-to-coast microwave relay “Long Lines” project, coincidentally enough also completed in 1951, a money-making operation.

By the mid-1950s, refinements of AMA, including Centralized Automatic Message Accounting (CAMA), which used things like banks of reed relays, ferrite cores, and cold-cathode tube amplifiers, began deployment nationwide.  The invention of the transistor in 1947, also by Bell Labs scientists, and the subsequent revolution in digital computing would pave the way for the eventual demise of AMA systems, but not for a while — it wasn’t until 1966 that magnetic tape started replacing the AMA paper tapes, and the concepts behind AMA and CAMA continued to tote up toll charge in computerized form until the mid-1970s.