The afternoon meeting was opened by Prof, F.C.Williams who presented A Storage System for use with Binary-Digital Computing Machines read before The Institution of Electrical Engineers on the 2nd November, 1948 and published in the Proceedings of the Institution Vol.96, Part III. p.81 (1949). The paper was originally submitted on 4th March 1948 and the published version was received on the 13 July, 1948. The work described was started by Professor Williams while he was still at the Telecommunications Research Establishment (T.R.E.) in Great Malvern, Worcs. and continued after his move to Manchester University.
The following summary is copied from the paper:
The requirement for digital computing machines of large storage capacity has led to the development of a storage system in which the digits are represented by a charge pattern on the screen of a cathode-ray tube. Initial tests have been confined to commercial tubes. Short-term memory of the order of 0.2 sec. is provided by the insulating properties of the screen material. Long-term memory is obtained by regenerating the charge pattern at a frequency greater than 5 c/s. The regeneration makes accurate stabilisation of the position of the charge pattern on the c.r. tube unnecessary.
The properties required of a storage system, and its operation as part of a machine, are stated. If such a machine were operated in the series mode, an instruction would be set up and obeyed in 600 µsec.
The main parts of the paper were:
The Conclusion of the paper was that It has been demonstrated that large numbers of digits can be stored on the screen of an ordinary commercial c.r. tube and that the development of special tubes for this purpose is worth pursuing and should lead to an increased storage capacity per c.r. tube.
The Discussion was opened by Mr. E.J.K.Hesketh (T.R.E.) who described a system of C.R.T. Storage for a parallel machine as follows:-
A C.R.T. storage system has been successfully built for operation in the T.R.E. parallel machine in which all the digits of a number of instruction are used simultaneously. A single C.R.T. will store all the digits of a given significance of 1024 different numbers. Twenty C.R.T.'s will be required for a twenty binary digit machine. On the other hand a complete operation on a number can be carried out in the time that is required to examine a single digit. In the present instance this is 20 micro-seconds; a similar time is required to assimilate an instruction.
The electronic problem has been solved of regenerating all the elements of a C.R.T. consecutively, while interleaving between these regenerations, periods when action can be taken at any point of the C.R.T. This is achieved by controlling the deflection voltages alternately by two registers, the regeneration register and an action register, each consisting of an array of trigger circuits. The regeneration register is a straight forward scale-of-ten counter. The configuration of the action register is determined by the computation itself and consists of the address of either an instruction to be assimilated, or of the number to be operated on.
The deflection system is required to reach a stable state before the C.R.T. beam is turned on. The equipment built is capable of driving 40 C.R.T.s in parallel and moving the beam right across the tube in 3 microseconds. In the experimental equipment a single C.R.T. has been made to store successfully when its deflection plates were connected to a 1500 pf condenser 40 times their own self capacity.
In reply to Mr. REY, Prof. WILLIAMS said that the CRT storage system was possibly not more compact than the ultrasonic delay store but it involved no precision engineering and was probably cheaper and more flexible. There was no period during which the stored information was not available and he saw no reason why the system should be less reliable. Mr. GOLD said that the mercury tanks now used in computing machines had originally been designed for other purposes and there was room for further development. There would, for example, be no difficulty in having several reading-out crystals spaced along the tube. Mercury tanks are robust and contain nothing that wears out. He considered that they were more reliable than cathode ray tubes or valves. Mr. WILKES said that a good deal of development had been done on mercury tanks in the United States. It was important to remember that quite a lot of electronic equipment was needed in an ultrasonic delay store in addition to the mercury tanks themselves. The largest item in cost and size was not in fact mercury and steel but the associated equipment.
Prof. WILLIAMS said they had not yet had sufficient experience to say whether careful selection of cathode ray tubes was necessary. Some of the desirable features were cleanliness of the glass, freedom from pinholes in the screen and good focus.
Dr. KOSTEN referred to the possibility of false signals occurring on account of bad spots on the surface of the tube and asked whether it would be possible to use two tubes together, rather in the same way that relay contacts are sometimes made double. Prof. WILLIAMS said that the use of two tubes in this way would not help, since if they gave different signals, there would be nothing to indicate which was wrong; it would be better to use three tubes. He did not think, however, that any such arrangement would be found necessary in practice.
The Methacon was mentioned as a possible storage device for a computing machine, Mr. ELLIOTT said that this tube had been described in the RCA review. It had two guns, one for reading and one for writing, and he did not think it would be suitable for use in a calculating machine.
