Computer Memories (continued)
The major elements of this project were based on queuing theory, a sequence of stored data or programs awaiting processing, and probability theory. We then took a modified "random walk" through that list of elements and events. In other words, we made a list of all the kinds of things that could or must happen, determined how often and in what sequence they might or must happen, and made them happen when and if it was necessary, or needed, to complete the battle, then determined the consequence of each action. The next event in the sequence was then modified by what had just happened.
The retired military officers and some of the active-duty officers were supposed to tell us what had happened when they were in battle in World War II and Korea, and what the Army said was to happen in such cases, from time to time, and what might actually happen, and the effect of that on an Army in battle.
One day, after our computer completed a couple of days of military exercises, the General looked at the data and said, "This can't be right, it just can't be right." The appropriate Ph.D. looked down his nose at the General, but he stood his ground and said, "I tried to move an Army over this exact part of Europe during WW II, and it took days and days, not hours."
After he insisted some more, the maps were re-examined. At a point where the contour lines were close together, the person entering data describing the terrain did not notice there was a high steep cliff, not just level ground. And that made a huge difference when we corrected the data, and re-ran this same problem.
Generally I like to remember I was involved in some of the earliest computer simulation projects, first at RAND Corp. in the mid-1950s, and at CEIR around 1960.
Since I had worked with computer simulation at the RAND Corporation (training courses for the Air-defense system), I was more familiar than most people with what we were trying to do at Ft. Huachuca. In this case, some of the higher-ups in our project generally "looked-down-their-nose" at the military, and paid little attention to the information the project gave the US Army. That made no sense, after all the US Army was paying for what we did.
Without being asked by the project director and not giving him a chance to say no, I wrote a Fortran program that converted the huge pile of number-covered paper that emerged from a game, into a set of words and statements that told what had happened. That did not endear me to the other people on this project, because now some of the military could actually understand a little of what we were trying to do!
BACK TO CALIFORNIA, AT RCA IN VAN NUYS
At RCA's Systems Division I worked on the computers for the Atlantic Missile Range, and the Saturn Missile Control System. One day my boss said he was going to a meeting where it was to be announced who was going to be laid off, and he would deliver my fate a few minutes later. The boss returned with a long, sad face, he had been laid-off, I still had a job - but for only a few more months.
MY JOB AT RCA
While working at RCA I gave sales presentations, conducted training courses, helped write computer manuals, programmer reference manuals, built a computer facility, and was involved in anything else they could think of.
I worked on special-purpose computers: the RCA 4101, 4102, and 4103, that were installed in a communication system for an airline, and on ships on the Atlantic Missile Range, among other places. There were rather basic differences between the three models, so you can imagine the potential confusion when I made presentations, or taught classes with users of all three systems, in the same room.
I also worked with the RCA 110A, the Saturn Ground Control System. This computer rode on the launch vehicle as it took the assembled rocket from the assembly building to the launch pad, at Cape Canaveral. It was used to monitor the Saturn Missiles up to the moment the engines were started and the astronauts on their way to the moon.
In those days there were no computer languages and operating system as there are today. I helped write the technical manuals that explained how the RCA computers worked, and Programmer Reference Manuals that told what the programmer needed to know to write the programs. I never thought to keep a copy, and now I have asked NASA and RCA, and no one has copies of those manuals anymore.
BUILDING A COMPUTER FACILITY
One time I was given the task of designing and overseeing the building of a computer facility to hold the RCA 4102 and the 110A computers.
In those days computers used a lot of electric power, and that generated so much heat that special air-conditioning was required. In addition, dozens of huge cables were needed to connect the various parts of the system, so the computer floor was constructed a foot above the regular floor, allowing both air-conditioning and cables to reach all the machines.
This floor consisted of panels that could be removed when it was necessary to change the location of tapes, drums, printers, etc., and the cables that connected them.
THE RCA 4100 SERIES COMPUTERS
I remember the address portion of an instruction on the RCA 4100 was 15 bits, so it could address 32,768 words of memory. Each bit of a certain 15 bit register indicated an interrupt device, and that bit would be "turned-on" when that device required attention.
Somehow that register was used as part of the addressing scheme, so if a device of a higher priority-- raising in priority from right to left -- demanded attention, the program was immediately transferred to that routine. The only reason this is mentioned with no more information than this, is because of all the computers I became familiar with, the interrupt system on the 4100 series computers, was unique, outstanding, and easy to use.
The RCA 4102 computer was used on ships on the Atlantic Ocean, under the path of the missiles, just after they were launched. The "picket ships" carried special RADAR equipment to track the missiles. These ships were floating, rolling, and pitching in the open ocean, and the RCA 4102 computer was used to compute the difference between the position that the ship was in when the RADAR signal was sent, and the ship's position when the signal came back.
You can imagine the concern we felt in Van Nuys one morning when, while listening to a radio description of a missile launch, we were informed that the launch had to be canceled because one of our computers had failed.
THE RCA 110A SATURN COMPUTER
The most interesting five or six weeks of my working career involved the Saturn Ground Control Computer, the RCA 110A. The RCA 110 computer was designed for such things as controlling machines at a steel company, and one was used to control commercials and programs at the NBC studios in Burbank.
While most computers can do many different kinds of jobs, in those days especially, some were intended to do payroll and accounting, some were designed to do scientific calculations, others were especially designed to respond to things that happened, then cause other things to happen.
With limitations in size and complexity of the job, and efficiency of the operation, all of these computers could most likely do any of those jobs.
NASA needed a process control computer quickly, and since this one existed, they ordered a few. Consider the large moving platform that is used to get the huge rockets from the buildings at Cape Canaveral, out to the launch pad: this computer was to ride on that platform, keeping that huge missile in tip-top condition, until it was ready to leave the ground.
The RCA 110 needed to be redesigned for NASA. In the original RCA 110, the 24 bit words were transferred from the memory stack to the memory register in a "bit-serial" mode, that is, one bit at a time. To increase the speed of operation in the RCA 110A, the 24 bit word was moved to the memory register in one transfer, with a resulting increase in speed. The instruction decoding methodology of both the 110 and the 110A required the instruction be decoded in the bit-serial mode.
The RCA 110A had a 24 binary bit word, the magnetic tape transferred data at 15 thousand characters per second, and the drum rotated at 3600 RPM. The typewriter operated at 10 characters per second (that's output, we couldn't type that fast), and the paper tape reader and punch, operated at 60 CPS. The following speeds are all in microseconds: Time to address and move a word from memory, 28.85; add/subtract time 57.7; multiply 779; divide 865.
I give those figures to illustrate the speed of the computer that controlled those thousands of lights, switches, clocks, etc., for the missile that delivered human beings to the moon.
Although I was well versed in the RCA 4102, all I knew about the RCA 110A was that it was painted gray, and the only one around was on the floor in little pieces, on the other side of the computer facility I was still in the process of building. The 110A was being rebuilt to NASA's specifications, and it didn't work yet. NASA in Huntsville, Alabama then decided they needed to be trained on that machine right now, so signed a contract with RCA for a training course.
A big problem arose about three weeks before the course was to start, when the only two men who had ever conducted training courses on that machine quit RCA. I filled in.
I had to learn about the existing machine and find out what was being redesigned and how that worked, create a three week training course, produce all the paper work, such as sample programs, tests, work sheets and all the rest that goes with such a thing. The only way to do this was to read the actual engineering blueprints.
While reading the blueprints of this computer, I discovered something that just could not work, and the lead engineer in charge of the redesign was out of town.
When I arrived at the NASA Huntsville office where the course was to be taught, with not one minute of advance notification, I was told they had decided that instead of training 20 men six or seven hours a day for three weeks, I was to speed it up and conduct two classes of 20 men each, one 5 hour class in the morning, and a 5 hour class in the afternoon. You can imagine how difficult it was to cover the information in less time, and then wonder, while presenting a point in the afternoon, if I had already said it a few minutes ago, or if I said it earlier that morning.