The Radiation Printer

George Michael

by George Michael

Editor's note: This is an expansion of an article that appeared in CORE 1.4,
a Computer Museum History Center publication, in November, 2000.
Used here with the permission of the Computer History Museum.

There are very few computer users left who still can recall the frustration of having to wait for a printout. For instance, around 1953-1954, at the Lawrence Livermore National Laboratory (LLNL), the first printers used in conjunction with the UNIVAC I--our first computer--were nothing more than typewriters with print rates of perhaps 6 characters per second. Since the typical output from a design calculation involved 50,000 to 100,000 characters, printing would take an inordinately long time. The quest for speedy printing at LLNL led us through a succession of interesting machines, one of which we relied on for about 10 years, starting in 1964. This was the so-called "Radiation Printer", an eccentric and demanding invention that met our computer printing needs for speed despite its own oddities.

To better appreciate what the Radiation Printer represented, it will be useful to consider first, the few printers that preceded it. I've already mentioned the use of suitably modified typewriters on the UNIVAC I. The IBM computers that arrived next had modified 407 accounting machines serving as printers. They were rated at 150 lines per minute, a line being up to 132 characters wide. They were also far too slow to meet our needs. As I recall things now, IBM never included faster printers for its initial computers. If you dealt with IBM, these 407s were what you got; a Hobson's choice.

One of our first attempts to get something faster than these printers arrived from Remington Rand around 1957. This was a 600-line-per-minute impact printer, where a line included any number of characters from 0 to 120; each page held about 50 lines. As fast as this was, it was still too slow to serve the needs of dozens of people who spent too much of their valuable time waiting for results. Also, when these so-called impact printers ran, the noise level was dangerously high A few intense users lost some of their hearing from standing in front of the printer, anxiously trying to read their output as it was being printed. In addition to being very noisy, impact printers were not sufficiently reliable, so we sought other solutions.

We tried a marriage of cathode ray tubes and xerography: The SC5000 built by Stromberg Carlson in 1959. This device formed characters by projecting an electron bean through a character mask, creating a spatial distribution of electrons then formed the selected character when plotted on the screen of the CRT. The SC5000 further selected where to position the character along the print line. The light thus generated was projected onto a selenium-coated drum that is fundamental in the xerographic printing procedure. In this process, after the image was formed on the selenium drum, it was dusted with xerographic powder ("toner"), which adhered only where the light had suitably charged the surface. By bringing paper in contact with the drum, the image was transferred. The paper then moved through an oven where the powder was fused to the paper. Input to the printing system was via magnetic tape.

One of the cabinets shown here is the SC5000, circa 1960, that was prone to catching the paper on fire. The print rate was one page per second, with input via magnetic tape. One can also see a homemade device for Z-folding the paper.

The SC5000s were modified so that they printed at an impressive rate of about one page per second. This required expanding the fusing oven and adding a Rube Goldberg device to z-fold the printed output. Quite often, the paper would catch fire as it moved through the fusing oven. The printer kept running, but now acting more like an automatic stoking device, feeding fresh paper into the fire! The SC5000 was very much like the angelic little girl with a curl right in the middle of her forehead: "when she was good, she was very, very good, but when she was bad, she was horrid."

Even when printouts were produced at the one-page-per-second rate, the total time was just too long to meet the aggregate needs of all users. The search for faster printing continued, so everyone was primed to welcome a new printing technology, ultimately embodied in the so-called "Radiation Printer".

Two technologies came together in the Radiation Printer. First, the actual print process was based on an electrographic printing technology, and second, the process was wedded to a standard printing press that far predated the advent of computers, but was rugged and reliable. Before the arrival of computers, most printing presses were designed to produce many copies of the same page. For LLNL applications on computers, the problem was to produce just one copy for each of thousands of output files. The electrographic technique, which is both fast and clean, uses light to carry information to an electrically charged material where a toner was used to make the image visible. The image is then transferred to paper where it is fixed by chemistry or heat. Xerography is a good example of this technology. Even though further discussion of the process is beyond the scope of this article, some basic differences as used in the Radiation Printer are important to note.

Mona Millings with the Radiation printer
Operator Mona Millings stands at the table where the separated output was delivered from the printer, and at the other end, the large rolls of paper used by the machine. Paper from the rolls could be spliced head to tail so there was no need for rethreading through the press. A roll lasted about 45 minutes and a special dolly was needed to move the rolls, since at over 200 pounds, they were far too heavy to be moved by hand.
Back Side of Printer
The lower photo shows the back side of the printer.

Instead of light, electronic charge was used to carry the information. The charge was made to produce an electric arc from a selected stylus to a black electrographic web through a whitened paint-like material that coated the web. The arc burned a tiny hole in the coating thereby revealing the blackness of the web. This made toning and fixing steps unnecessary. One saw a black dot, and enough black dots produced a simulacrum of the image sent by the computer.

This type of printing process was normally used for the production of mailing labels for magazines like Time and Newsweek. Although no actual printer existed, everyone felt confident that a printer could be scaled up from a mailing label size to a larger page format, and it seemed it could be made to go quite fast and it promised to be economical. We solicited bids for a high-speed printer, and what became known as the Radiation Printer was chosen.

Some salesman got the Radiation Printer to brag about itself. Here is a quote of some of what it said (I omitted some parts that were inaccurate):

The printer had 600 styli arranged in 100-styli modules. The print area was about eleven inches in width, and the page was eleven inches tall. The images were not considered to be very high resolution. A traditional printing press was used to move the web past the styli. The procedure was dubbed "Revelation Printing," because the coating was burned away by the charge coming from the styli, thus revealing the black paper underneath.

An annoying problem with the styli arose during operation; they tended to get contaminated with burnt paint debris, and therefore stopped functioning. The solution had nothing to do with modern technology. Cleanliness was achieved by blowing pulverized walnut shells against the styli. It was claimed that other nuts would not work.

A few additional remarks seem to be in order. First, the Radiation Printer had nothing to do with radiation, but simply was named for the company that built the printer: Radiation, Inc., of Melbourne, Florida. The company modified a real (Hamilton) printing press and added the needed electronics and controls to produce a printer that ran at seven pages per second (for Indy drivers, this turns out to be about 4.3 mph.) Printers in the newspaper business run even faster although they don't seem as versatile. In addition to printing at that speed, it punched binding holes at the top and bottom of each page, perforated each page so jobs could be separated, fan-folded the output, and separated the jobs one from another. The various performance numbers for the printer are summarized in the tables below.

There were enough styli to allow up to 120 character positions per print line, and each character was formed within a 7 x 9 dot matrix. Suitable spacing between characters and between lines of characters was thereby provided, so that in practice, a page could contain up to 10 columns of numbers each up to 12 decimal digits, each column containing 55 to 60 numbers. The capacity of a page was thus about 5,000 characters. It was also possible (but not easy) to address any point in a line, so that with some special programming tricks, graphs could be produced. Printing was thus accomplished exactly as a video-scanned raster is produced.

Something in the print process gave the output a disagreeable odor. Some of the users actually complained of headaches. An investigation of the odors failed to expose any serious health hazards, so the simplest response to this was to authorize the use of fans that could keep the odors away from those sensitive noses.

Radiation printer
The machine page-perforated, hole-punched, fan-folded, separated the print jobs, and deposited them onto a slowly moving set of belts.

So what does seven pages per second mean to the users? Each page was approximately 11 inches square. This implies the speed of the paper through the printer was about 77 inches per second. The print data was supplied from any magnetic tape able to provide a nominal 60,000 characters per second--we used IBM 729 tape handlers written at 800 characters per inch. Such tapes had a nominal rate of transfer of up to 62,500 characters per second, more than adequate for printing, so the extra time available allowed for the filling and emptying of buffers, and for the movement of the paper past areas at the top and bottom where no printing was done. On balance then, of the 7 pages per second, about 1.3 pages-worth of that time was not used for printing, but for the extra movement of paper required to get from one page to the next, as well as time for hole punching and page scoring. Thus the rated speed of 7 pages per second meant that the user was getting about 6 completed pages per second within that 7-page time. This printer minimized user waiting time or it serviced more users in a given interval. Very few of the users (physicists) could read at this speed. In effect, then, the throughput speed of this printer was generally adequate to meet the needs of the growing user community, and it did so for a bit over ten years.

The Radiation Printer was integrated into the normal operations of the computation department, and very quickly was producing around forty million pages per year. This was only about one-fifth to one-third of its capacity, which was a good thing. The machine could be taken down for emergency maintenance, and still very quickly clean out the entire print backlog when it was brought back on line. Later on during its tenure, some microfiche recorders were added. Their annual output quickly grew to about 130 million pages distributed over about one million pieces of fiche. The effect on the Radiation Printer was less than expected however: The annual output dropped to around thirty million pages per year and stayed there. For most users, the fiche was used for long-term storage of their computational results, and output from the Radiation Printer was used mostly for day-to-day checking. When a project was finished, the paper was generally discarded.

Paper for the Radiation printer
Rolls of paper waiting to be fed into the Radiation Printer

The output from the Radiation Printer was hard to read; the gray-on-black paper was heavier than ordinary paper; it had, for some, an undesirable odor; and it took up too much storage space. The output was not pretty; the users often referred to the printouts as "scunge," but it met their needs, producing at the rate of 7 pages per second. None of the printers that were brought in to replace it ever came close to this speed. However, as effective as the printer was, no one shed a tear when it was removed sometime during the late 1970s.

It's always humbling and sometimes instructive to ask if anything was learned. There are several lessons available, though who learned them is not clear, nor is the question of whether the lessons have had any long-term positive effects. Somewhat in the spirit of a post mortem here are some things that were learnable:

In the course of dealing with users of all sorts, we evolved an additional rule to help get through the day: Generally, if somebody doesn't know what to do, don't ask them.

Radiation printer output
The paper required for the Radiation printer was a sandwich of a black conductive layer coated with white top layer. The overall appearance was a blueish gray. Printing was accomplished by an electric arc burning a tiny hole in the white coating to reveal the black layer underneath. Too bad we can't reproduce the odor here!

Table 1: Approximate pages of computer printouts
per month in 1978
Teletypes 200,000
35 mm film 600,000
On line Printers 830,000
6 Microfiche Recorders 9,800,000
Radiation Printer 3,400,000

Table 2: Early computer printing to 1974 (Approximate speeds)
Typewriters 1953 0.5 Lines/sec
Line Printers (IBM 407) 1954 2.5 Lines/sec
High Speed Printer (Rem. Rand) 1958 10 Lines/sec
SC5000 1959 60 Lines/sec
Radiation Printer 1964 420 Lines/sec

Table 3. A Summary of Radiation Printer Performance Numbers.
Print Technology Electrographic, Revelation
Data source Magnetic Tape, up to 800 bpi; 75 ips.
Character Rate up to 62.5 Kcps
Print Rate 7 pages/sec; 4.3 mph
Print size 5000ch/page

A NOTE ABOUT DATES: More precise dates may exist, but most official records are in a state of flux. The dates used here are my best approximations. Similarly, the values in the Tables are extracted partly from several unpublished internal reports. They are intended mostly for comparisons.