CSIRAC had two main storage systems. These could be described as the RAM and the Hard Disk in todays terms.
Brian Cooper, a third senior member of the design team, designed and built the magnetic “drum” storage system. The design work on it started in 1952, and a number of trials of different fast and slow designs were performed, including a drum-based device of 1024-word capacity (disk drives of the 1990’s with a capacity of 1000000 bytes are used in entry-level personal computers) and a mean access time of only 5-milliseconds (disk drives today can only just match that speed!).
By 1956, a horizontal-axis disk-type device was permanently installed with one segment of 1024 words in use. This store was a four-segment, 20-bit parallel system (it had no seek time, as it read all 20 bits in parallel; the access time was due to the rotational latency of the spinning disk), each segment being addressed by the 10-bit address of an instruction (there was no Disk Operating System, as the data was directly addressable). Each segment was given its appropriate source or destination code, and a 10-bit coincidence between this address and the disk clock track count was detected as the drum rotated and serial-parallel transformations and time inter-clocking performed.
CSIRAC’s main memory design allowed for thirty-two acoustic mercury delay lines, each with the capacity to store 16 words (of twenty bits); this was later upgraded to 32 words when a method of interleaving was devised by Reginald Ryan. A total of 1024 words of storage was therefore possible but only a maximum of 768 were ever available!
Mercury delay line technology was developed for radar systems during World War II. These worked by storing data as a series of acoustic pulses in a tube of mercury.
Some of the CSIRAC mercury tubes were about 10mm in diameter and 150cm long. A modulated pulse was generated by a transducer at one end of the tube, and 960 microseconds later it arrived at the other end of the tube where it was received by another transducer. The modulated pulse was detected, amplified, re-shaped and re-generated, being fed back into the beginning of the tube.
The tubes were housed in a box affectionately referred to as the coffin. This was a wooden box about the size of a real coffin and contained a heater which kept the tubes at a constant temperature of about 40 degrees Celsius (about the hottest Melbourne summer’s day!).
All of the incidental registers were also mercury delay lines, however they were only about 6 inches long and were housed in another box above one of the control racks.