My Career at RadiationFrank Perkins

Radicon/Radiplex

I was hired at Radiation (on November 9, 1959) to shepherd a family of standard products, including the Radicon analog-to-digital converter, and the Radiplex analog multiplexer. Almost every system Radiation built included these components and someone had decided it would be a good idea to have utilize standard products in these systems. The units received a full product design treatment from an outside consultant, and were physically quite striking, with curved, gold anodized front panels. Electrically, the units had been breadboarded, but had never been tested as complete units, particularly to their full specifications. The designers were all busy working on new systems and were not available to support the standard products. It was my job to turn this raw work in progress into a useful product. To further complicate the issue, the “standard” Radicon was designed to operate in many modes and configurations, and was therefore very complicated (and expensive).
The units had been designed into a number of systems under development, but all of the project engineers had insisted on simpler configurations customized to their particular requirements. Furthermore, new semiconductor components were becoming available which allowed much more efficient implementation of some functions such as analog switches. The original Radicon used matched quads of diodes for analog switches. These required large reference voltages to make the switches sufficiently accurate. The Radicon regulated a 200 volt supply down to 150 volts for this purpose. The resulting power dissipation caused the regulator card to run so hot that you could burn your hand on the card edge. Later versions used transistor switches, in turn allowing a 8 volt reference with much lower power dissipation.
Eventually I managed to get the advances incorporated into a simplified version of the “standard” Radicon, and to make other changes, including use of a tunnel diode error amp to get this unit performing to full 12 bit resolution, but this was not easy.


One interesting aspects of the Radicon program was that I had a good excuse to talk to with all of the project engineers using the Radicon in their systems. This provided exposure to the full array of interesting systems Radiation was building.
One of these systems was the “Pink Kitchen”, for General Electric. This was a set of what we would now call computer peripherals, such as format converters, tape drives, high speed printer/plotters, A/D and D/A converters, etc. These were designed, like the computer mainframes of the day, to be displayed in a glass-windowed room to impress visitors, hence the stylish design and striking color scheme.
Another interesting system, built for White Sands Missile range, used a “hybrid” version of the Radicon to allow automatic scaling of multiplexed analog signals as they were digitized. This was desirable because multiplications were so time consuming in digital computers of the day. (The Radicon accomplished this multiplication by multiplexing the reference voltage in sync with the signal multiplexing, thus scaling the signal.)
Another notable system was for Norair (ex Northrup Aircraft Company). This system was used in flight testing the Norair jet trainer which subsequently became the T-38, used to this day (50 years later) by the shuttle astronauts. Working on this system was Bernard France, who became a close friend and lunch companion for many years until he recently moved to Washington state.
The largest single user of Radicons were the data acquisition systems for Thikol Corporation, in Utah, for development of the solid rocket motors for the Minuteman missile. These systems were also heavy uses of the Radiplex 89, the first low level multiplexer, which eliminated a number of complex amplifiers.
The absence of multiple precision amplifiers allowed for a more compact data acquisition system, and the idea of a portable cart-mounted data acquisition system attracted the attention of a California subsidiary, Radiation at Stanford. It seemed simple enough, but the development was plagued with problems and delays, caused by inexperience in designing such systems, confusion in the specifications of the hardware and the immaturity of the hardware. I remember a number of emergency trips I made to the Palo Alto area to solve problems in applying the Radicon and Radiplex. In those days the only jet service to California was from Chicago, so I was always faced with a plane change in the Chicago winter, before the days of jetways. With balmy weather at both ends, I hated to wear an overcoat just for the transfer, but it was really cold on the outdoor run to the airplane in Chicago. After a few of these emergency weekend trips, I discovered that I could upgrade to first class in Chicago with my company Air Travel Card and felt fully justified in so doing. No one at Radiation ever questioned this expenditure.
Eventually the California division ran out of money for this project and lost interest, and the unfinished system was brought to Melbourne and reassigned to the Products Division.

Tape Format Converter

As all this was going on I was assigned to work on the Boeing Tape Format Converter (TFC), intended for use in their development of the Minuteman ICBM. The TFC had an Ampex serial tape recorder, which played back the recorded PCM telemetry bit stream for recording on computer compatible parallel tape on a pair of Ampex FR400's. The pair of FR400's was necessary to allow continuous operation of the serial tape recorder. My job was the design of the control panel, largely devoted managing the digital tape recorders. This required individually mounting and prepping the tapes, handling the automatic or manual switching between them, and automatically terminating the tape to suit the specifications of the computer.
I always regarded the TFC job, under the leadership of Larry Klingler, as an example of how electronic design should be done. Larry was a very smart individual, fun to work with, but with strong ideas on how things should be done. One of his pet peeves was control panel design. He strongly felt that system operation should be intuitive and that the design should minimize the chance of an operator error. I can still feel Larry roll over in his grave at the poor design of Personal Computer controls; he would never tolerate clicking the “Start” button to stop the computer.
His methodology as a project engineer was well thought out. He verbally described to me the functions to be performed with the control panel and then cut me loose. He regularly, but informally reviewed the design with me, having me describe what the operator did, and how the system responded. When the system was finished I felt very flattered when he complimented me on the design and operation of my part of it.
Subsequent formal efforts at design reviews degenerated into farces. A detailed written report would be generated and circulated to a panel of “experts”. Then their would be a big meeting with vu-graphs and handouts. This was all a big drag on the time and effort of the designer and by the time it all took place it was dreadfully late to make any design changes. I remember many times making a
suggestion in a design review meeting and being told “that's a good idea, but it is too late implement it.” That made it seem futile to have the big presentation if it were too late to take advantage of suggestions. Truthfully, the objective was to check off the box that a design review had been held. In contrast, Larry's informal reviews made sure that his experience was incorporated in the design, but did not require a lot of time-wasting documentation.



Huntsville Sync Study

One significant event was my assignment to a PCM synchronization study that Radiation won from NASA-Huntsville. I don't remember being involved in the proposal or marketing for this program, nor exactly why I was assigned to it, but it had a notable impact on my career. For one thing, I became somewhat of the local guru in synchronization at Radiation. It also marked my first extensive collaboration with Dan McRae an analytical genius in our Advanced technology department. This collaboration continued throughout my career at Radiation and Harris.
PCM synchronization was a immature science in those days. The NRZ format was the most bandwidth-efficient but had no discrete frequency component for bit synchronization. At least partially because of this concern, the Minuteman telemetry devoted 3 bits of 27 to a word sync pattern, for a 11% performance penalty. Someone had recognized that passing the NRZ format through a non-linearity would create a bit rate component for synchronization, but no one knew if this would provide robust enough synchronization. This was one of the questions to be answered by the Huntsville study. I breadboarded a number of synchronization approaches and showed that bit rate synchronization need not limit PCM performance. (It should be noted that the present day Internet suffers a 10% performance penalty through the use of a character oriented format, which can be viewed as a from of word synchronization.)
Frame synchronization was also an issue. People generally understood that the insertion of a periodic sync pattern in the data was a good approach to frame sync, but it was unclear what parameters of pattern length and frame rate were appropriate, and what exact sync algorithms would yield satisfactory performance. Dan devised analytical procedures which allowed analysis of frame sync performance in terms of format parameters and synchronization strategy and settings.
The insights into bit and frame synchronization served Radiation well in winning a number of PCM Decom system jobs, which were a significant business area for us at this point in time.

Bit and frame Synchronizers

I designed and built several bit and frame (group) synchronizers for various decom jobs. The bit synchronizer was particularly important because it affected the overall performance of the system.
The bit synchronizer (or Signal Conditioner) accepted a noisy analog version of the PCM bit stream (from a radio receiver, for example) and output a binary (digital) bit stream and an associated clock signal. They were a challenging device because they had to operate over a very wide bit rate range (typically one bps to one million bps) and needed to synchronize and make bit decisions in a near optimum manner in the face of noise, jitter and other disturbances. I built a few synchronizers of PC cards (see “Cards vs Modules” in the Appendix), but modules offered the chance to build a compact modular synchronizer that was dubbed the Model 5220. My first version had a design flaw in the form of leaky selection switches for the loop filter, which caused flaky behavior under some conditions.
These synchronizers happened to be for NASA-Goddard, who were often a difficult customer and this time happened to have a particularly antagonistic engineer on the job. He found the main problem in the units, but proceeded to nit-pick with a bunch of other mostly made up problems. He and I grated so badly that Radiation management wisely kept us from talking while I fixed the real problems. Eventually this NASA engineer came to Radiation to present his case for junking the 5220. I wasn't allowed to attend the meeting, but Dan McRae and others demolished his technical arguments and the updated 5220 went on to wide and successful service.



PCM Decoms

A PCM Decom accepts a serial stream and routes each channel to a particular output channel or device, such as a D/A converter. PCM formats were flexible and could be quite complex (see Appendix “”PCM Telemetry”). (The Gemini format even changed in flight.) Decoms need to be flexible to handle the variety of formats. Early decoms, such as one we built of McDonnell Corp. for the Gemini program, achieved this flexibility with patch panels, but the programming of these with patch cords was a complicated nightmare. A better approach was to have electronic storage of all the parameters and have the decom look them up as it need them. This was easier said than done with the components available in the 1960's, considering that the short time available for processing a word. A system built for North American Aviation used a rudimentary stored program decom, but the first system to really exploit the benefits of a stored program approach was the 540 Decom designed for NASA. It was used for testing and checking out the various systems used in the Apollo moon landing program. The procuring agency was the Manned Spaceflight Center at Houston. The marketing of the systems involved lots of maneuvering by enthusiastic marketing types, hiring away competitors engineering talent, and late night proposal writing by a small group of Radiation engineers, including me. The final result was a sophisticated system that performed its tasks well and as widely used at a number of plants and sites.
Bernard France devised an efficient technique for performing the necessary . On the early versions I mostly remember the long hours of checkout, trying to get the stored program technology working, under severe cost and schedule pressure. In spite of the pressure it was fun because of the other engineers working on it, and the relaxed but effective management of Al Medler. Al would come in to be with us on the long night shifts and was comforting support even when he dozed off on the stool. We finally met all the spec and were rewarded with what to us a “production” contact for 20 or so systems ,installed at Cape Canaveral, Houston (MSFC), Daytona Beach (GE), and Downey, California (NAA plant). In the later phases of the program I was the senior technical person working on the job and got to adjust a few details to suit my preferences. I wasn't the titular system engineer, but I got to do most of the fun management under Medler's gentle guidance. One fun aspect was that we were an autonomous organization, with our own leased building. We also had a great staff of engineers. I particularly remember Bob Whitlow, Gerry Duggan, Jim Pettigrew and Jack Wark.
We built a number of other stored program decoms. The PCM-DHE, for NASA's worldwide tracking network, was built around the same time as the 540's, using the same technology.
Later we built the SGLS (Space-Ground Link System) decoms for the Air Force.



In 1967, things started to change with the acquisition of Radiation by Harris-Intertype of Cleveland, Ohio, later to become Harris Corporation.