By Darwyn F. Kelley
SOLID STATE TECHNOLOGY
In 1952, Univac engineers developed a magnetic amplifier which they later trademarked under the name of
Ferractor. Each of these tiny components could do the work of a vacuum tube and do it better with greater reliability. Their operating life was much longer.
The obvious next step was to build a computer utilizing Ferractor magnetic amplifiers instead or using vacuum tubes. In
1956 such a computer, the first of its kind, was delivered to the Air Research Center at Cambridge, Mass.
A Solid State Computer naturally was much easier to maintain then the older vacuum computers. The technique of using Ferractor Amplifiers, together with miniaturized circuit elements such as resistors and capacitors were mounted on printed circuit boards . This was a revolutionary system. Two or three of these printed circuit boards replaced vacuum tube chassis. Instead of tracing a fault within the chassis, one would simply remove the faulty board from the rack and replace it with a new one. A simple procedure today but bear in mind we are talking in the 1950-1960 time frames.
The commercial version of this System was called the UVIVAC Solid State
Computer. The first one installed in the Dresdner Bank in Hanover Germany, in
1956. This was the first delivery of an all solid state computer, to a commercial user anywhere in the world. Over 300 were installed and this writer was the lead Univac System Representative at the Bethlehem Steel Works in Baltimore, MD.
(1960) The Solid State Computer was very interesting and fun to program.
The main memory in the Solid State Computer was a 5000 word rotating drum. Generally high speed drums were very efficient and reliable but had one major flaw, one had to consider rotating latency when searching this device. You had to plan on when the drum, with your data , was under the read write heads. This latency problem did slow down your access time. A better solution was
In the 1950's there was such a device in the laboratory stage of development. This was coincident-current memory, made of tiny solid state toroids, physically something like the Ferractor Magnetic Amplifiers, but with different properties. These toroids could be magnetized in one of two directions, and the direction of magnetization could be used logically in the computer circuitry to represent the on or off-or the one or zero-state of the first vacuum tube memories. Reversal of the direction of magnetization is accomplished by means of electrical pulses transmitted over wires threaded through the toroids.
Core storage worked so well it became main storage for the computer itself. The
UNIVAC 1103A Scientific Computer was the first computer to have such a memory. The
UNIVAC 1107 Computer was the first to use thin film storage for the ultra-high speed portion of its memory.
NEW CONCEPTS IN LOGICAL DESIGN
The discussion so far has dealt largely with improvements in the hardware, but there are other approaches to increasing the efficiency and general usefulness of a computer. One is to take a look at the logic of the machine itself, and this is what was done with the
LARC is the first system to employ what was called multi-level logic. That is, instead of thinking of the system as device in which data enters the memory through input and is acted upon serially, with the results becoming available serially through output media, the
designers thought of a system where many things could be happening simultaneously, even including the running of more than
one program at a time.
LARC actually used a second computer, called the Processor, to control traffic in and out of the main computer, the Arithmetic Unit. This freed the Arithmetic Unit for useful work 90 per cent of the time.
Even the secondary LARC Computer, the Processor, is freed of many housekeeping chores by the use of input-output synchronizers for control of the peripheral equipment. Since the Processor has arithmetic circuitry, it is feasible to run a business data-processing problem on the Processor while the Arithmetic Unit, that is, the main computer, is working on a lengthy scientific problem. There was no physical limitations on the number of input-output channels available.
The LARC core storage could store up to 97,500 computer words. More important than its size, however is the fact that a memory is modular and that the modules may be accessed independently.
LARC could do all these things simultaneously: Compute, accept information from the console, read from two tapes, write on two tapes, read from three drums, write on two drums, read 450 cards per minute, operate two high speed printers on line, and operate two electronic page recorders on line.
LARC was more than a fantastic piece of Hardware. A the time it was built and performance-tested, it could handle systems of equations of a magnitude beyond the reach of any other computer.
Two of the systems were delivered in 1960, the first at the Lawrence Radiation Laboratory,
operated by the University of California for the Atomic Energy Commission and the second, to the Navy's David Taylor Model Basin, Washington, DC.
In passing the acceptance tests required by the Laboratory and the Model Basin, LARC,
turned in some remarkable records. In the first performance test in April 1960, the Arithmetic Unit performed about 28 billion operations in two days. In every respect, the system performed beyond the demands of all the most stringent test ever devised for a computer. The severity of the test was due in part to the fact that the Lawrence Radiation Lab people who devised it were already among the most sophisticated computer users in the business.