ARRA II

Development for the ARRA II started at the Mathematical Center in 1952. After working on the ARRA I (until then still called just ARRA) Scholten and Loopstra joined forces with Blaauw on the first of November, 1952, to create a brand new machine called the ARRA II. This name was chosen merely for political and financial reasons, technically the two machines were completely separate. Any visitor of the MC would easily see through the official name, as the two machines were setup clearly separated, with not a single wire connecting the two contraptions.

ARRA I and II in one room

The ARRA I and the console of ARRA II in one room. Not a single wire connected the two computers.

Clock

The first ARRA machine also used "operation completed" signals to trigger new operations. For the ARRA II Blaauw introduced a clocked machine[1]. Just like the first machine the new ARRA used drum memory, and a tape reader for input. Its output was delivered through an electrical typewriter. 16 Of its keys were controllable by placing magnets under them. Plug connections were also used so that faulty units could more easily be replaced.

The And/Or ports were built using selenium diodes. These diodes came in the shape of rectifiers, but were stripped apart to retrieve the internal rectifier boards. These had a diameter of just roughly half a dime, whereas the complete rectifiers were large to huge elements. Because selenium could not be soldered - they simply couldn't withstand the high temperatures - the diodes were placed in an isolating FR-2 (Flame Resistant 2) board. The same material is used for printed circuit boards. In this board holes were drilled, with metal bits on each end. With the selenium parts in between, together with a compression spring, the metal tops could be connected, rather than the selenium directly. To connect these parts silver paint was used. Unfortunately these connections were highly unreliable, and the selenium was later replaced with germanium diodes, which could be soldered without problems[2].

The ARRA II had a two-register, fixed point serial arithmetic unit, 1024 words memory, divided over 32 tracks with 32 words each. Two instructions were stored in each word. Its instruction code was symmetric and consisted of 25 instructions. Six of them involved only the A-register, with a similar set for the S-register. Conversion also was no longer a special instruction. Instead two fast multiplications by 10 were provided by the machine.

Dijkstra did a lot of programming for the ARRA II. He also re-implemented x := x + 1 to be less prone to failure. This saved many maintenance calls at night.

Echo Check Routine

Dijkstra also built in an echo check routine. At the same time when a number was printed, the same number was being rebuilt using the decimal to binary conversion and back. This check routine was used in the ARRA II and its three successors. Due to the unreliable drum memory the "x := x + 1" operation was also rewritten to the following procedure[3]:

repeat A:= x; A:= A + 1; y:= A;
       A := -x; A:= A + y; A := A - 1 until A = 0;
repeat A:= y; x:= A; A:= -y; A:= A + x until A = 0;

Although more complex, it was far more reliable. With the machine often being used at night as well, unattended, this new procedure resulted in far less service calls during the night.

Blaauw working at ARRA II

Gerrit Blaauw working at the ARRA II.

Usage

As Dijkstra recollects Blaauw was very devoted to the job of completing the ARRA II. He also gave a lot of technical input, introduced proper documentation techniques, and other technical features of the machine. Nevertheless Blaauw left the MC fairly quickly again, just after a few years. He just didn't fit in too naturally, according to Dijkstra.

Dijkstra once made a long trip to Australia, of which the last part was by Fokker F27 aircraft. Upon his arrival his host was very apologetic that this world traveler had to do the last leg of the journey on such a shaky two-engine turboprop. Dijkstra responded: "Dr. Stanton, I felt quite safe. I calculated the resonance frequencies of the wings myself." [4]

The machine was first used in 1953. In December of that year it was first starting to be used. Its development took only 13 months, although in the first months of usage it was still being improved. A copy of the machine was made, with the name of FERTA, developed specifically for Dutch air plane manufacturer Fokker. It only had some slight modifications compared to the ARRA II, mainly some more elaborate shift instructions. One of the main characteristics of the ARRA II was its speed: it only required just over 20 milliseconds, the time the drum required to make one rotation, to calculate an addition. Multiplications and divisions took about 5 times as long. In 1956 the ARRA II was ceased to be used, but until then it was used continuously[5].


References

  1. Dijkstra, E.W. , "A programmer's early memories", A History of Computing in the Twentieth Century: a Collection of Essays: Academic Press, pp. 563-573, 1980  .
  2. Scholten, C.S. , Computers Ontwerpen, Toen, , 1979  .
  3. Dijkstra, E.W. , "A programmer's early memories", A History of Computing in the Twentieth Century: a Collection of Essays: Academic Press, pp. 563-573, 1980  .
  4. Philip L. Frana , Oral history interview with Edsger W. Dijkstra, , 2001  .
  5. Alberts, G., Blij, F. van der, Nuis, J. , Zij Mogen Uiteraard Daarbij De Zuivere Wiskunde Niet Verwaarloozen, , Amsterdam, Centrum voor Wiskunde en Informatica (CWI), 1987  .
Pictures: 
Previous Image 1/3 NextARRA I and ARRA IIARRA I and ARRA IIThe ARRA I and II shared the room. That and their names were all they had in common too.