The Typesetter that's part TV set: the Linotron 303
|This article describes some complex and challenging equipment I serviced years ago. I dealt with a wide range of service issues, even providing fixes for hardware and software issues of a 16-bit minicomputer.|
A character grid. These images are selectively televised and reproduced on film.
The Flying Spot Scanner - essentially a specialized TV camera for transparencies.
The film transport and the carriage-mounted CRT that reproduce the images onto film.
Front view of a 303. The Prime 100 minicomputer is top and center; the film transport is to the right.
Rear view of a 303, with the Flying Spot Scanner assembly visible below (CRT at left; photomultiplier at right).
One of the "high tension" CRT power supplies.
One of the seven main circuit boards. This one generates X and Y sawtooth sweeps for the CRTs, with resolution and amplitude determined by digital commands via the computer interface.
The Linotron 303/TC is a typesetter made by the Mergenthaler Linotype Company and used by a corporate client of mine from the late 1970's into the 1990's. The range of technology the 303 employed to get the job done was truly remarkable, including digital computers, analog computers, cathode ray tubes, photographic film, photomultipliers, video, some lenses and stepper motors and dozens of gears and sensors. The 303 guaranteed an extensive hands-on education for the person servicing it.
The purpose of a typesetting machine is to render an electronic document into a visual image; ie, hard copy, using whatever typefaces and so on are specified. The 303 accomplished the task using a television approach. It had inside it miniature black-and-white (actually black-and-clear) transparencies of all the necessary letters, digits and punctuation symbols, and it directed its TV "camera" to televise these characters one after the other to match the sequence in the document. As each character was scanned its image was simultaneously reproduced by a lens and a tiny cathode ray tube (CRT) moving past photographic film. When the film was developed, there was the document!
the flying spot scanner
The 303's TV "camera" was in the form of a flying spot scanner, a device that scanned the miniature transparencies by means of a moving point-source of light on one side and a photo-detector on the other. In this scheme a large CRT provides the illumination. The dot of light on its face moves in a raster pattern to progressively illuminate microscopic points on the desired character. As the illumination point passed over clear and opaque areas of the character, the photomultiplier sensed the resulting intervals of light and shadow. The light-vs.-shadow signal; ie video, ultimately got passed to the little output CRT, which would be scanning in unison as it moved past the film.
A stepper motor powered the carriage that moved the little CRT horizontally across the film, and another stepper motor advanced the film in the vertical axis after each line was done. Controlling the 303 was a Prime 100, a 16-bit minicomputer endowed with a humble but adequate 16K of ram. Digitally controlled oscillators provided rampable pulse trains to the stepper motors; the oscillators used binary-weighted networks of discreet resistors switched by FET's driven by TTL outputs from the computer interface. A similar scheme generated the sawtooth waves for the CRT deflection circuits.
The deflection signals got extensive analog processing, both for geometry correction and to adjust the X and Y scanning amplitude and thus the height and width of the reproduced character. (Italic script was achieved by bleeding a little bit of the Y sawtooth into the X deflection.) Push-pull power amps drove the CRT deflection coils, and finicky, regulated, high-voltage supplies were provided to drive the CRT anodes. This is actually a simplified description; it was quite a complex system.
My client had three 303's in operation, and you can be sure we had our share of malfunctions over the years. Sometimes the problem was just a speck of dust in the optics. And, predictably enough, we had more than a few bad connections in the hundreds of interconnects between circuit boards. But once in a while there'd be a more interesting problem, like a semiconductor failure in the masses of 7400-series TTL or in the sensitive analog circuitry and its power supplies.
the PRIME 100 Computer
The Prime 100 minicomputer was another whole world of detail in itself. The software took its input stream from an IBM 1130 in the department and proceeded to calculate all the character spacings and so forth. Then Prime 100 had the task of "spelling out the details" to the 303 via the 303 interface.
the computer interface
The interface had a TTL output register for each parameter. Some registers were 8 or 10 bits wide; others just 2 or 3. Altogether it was a 59-bit interface, specifying nine parameters. For speed, most of the registers were double buffered: as the 303 was executing one command, the computer would be busy setting up the next one in the pipeline.
hacking the Prime 100 code
The program for the Prime computer was stored on paper tape, and booting up the typesetter entailed feeding a little roll of tape into a tape-reader. There was also an old ASR teletype hooked up to the Prime, and after booting up you'd have to hit a couple of characters on the ASR to choose your input option. Unfortunately in our case the option was the same every time, so really the ASR was just a waste of space and a burden to maintain. Eventually we got to put the darn thing out to pasture. I reverse-engineered the Prime object-code on the paper tape and created a modified version that bypassed the superfluous ASR routine! The client was pleased with that.
an audacious repair
They were also pleased the time I managed to repair a Prime CPU board that'd failed! That CPU was a Board From Hell if ever I saw one — about 200 SSI and MSI TTL chips crammed chock-a-block onto a 16" by 18", eight-layer (or so) PCB. I had a schematic, but it wasn't for that exact board, leading to some unwelcome guesswork. Worse yet, there was no extender card, meaning that I had to do all my testing with the board still plugged into the backplane and card cage. Nevertheless, I managed to discover that a line in one of the internal 16-bit buses was stuck at logic low. Replacing a 74151 multiplexer solved the problem.