Electrostatic 'Memory' System for Digital Computing Machines.
The storage tube looks very much like on ordinary oscilloscope cathode ray tube.
Digital computers need some form of storage mechanism to enable solutions of problems and instructions concerning succeeding operations to be retained. There are various means by which this can be achieved, including mercury delay lines, magnetic recording and electrical charge storage on an insulated surface. The last-named method was chosen for the digital computer now in use at Manchester University and, in particular, cathode ray tubes were used as the storage medium.
The initial hopes of using ordinary commercially made cathode ray tubes as the storage medium did not materialize. When tried, such tubes were found to have two main defects: (a) random spurious pulses over the screen area (known as 'phoneys'), and (b) small total storage capacity. A special tube, shown above, was therefore developed at the GEC Research Laboratories to overcome these disadvantages. This is an electrostatic-deflection tube incorporating an electrostatically focused gun to give a very small spot and a fluorescent-screen storage surface.
In operation, the computer uses the binary scale and the two digits 0 and 1 are represented by charge distributions on the storage surface. These different charge distributions can be established by various techniques, for example, by the spot being either defocused or focused, commonly known as focus, or by the focused spot being drawn out into a small line, known as the dot-dash system. Although the last-mentioned system was used originally, the reliability of the storage surface was impaired if there were any 'phoneys' of amplitude greater than 50 per cent of the initial pulse at the beginning of the trace. The defocus-focus system is now used, and with this it is possible to operate a tube having a much larger amplitude of 'phoney' signal.
The display of these binary digits can be seen on the face of the storage tube since the electron beam is intensity modulated as it scans over a rectangular raster. A dot pattern is thus produced, a typical pattern consisting of 64 rows of 32 dots per row.
As the beam scans over the previously deposited charges on the storage surface, current waveforms are capacitively induced in a 'pick-up' plate which is fixed to the outside surface of the storage tube screen.
Showing how positive and negative pulses are strobed to represent binary digits.
These waveforms are shown above. By strobing, or selecting a particular portion of these waveforms, positive and negative pulses can be made to represent the two binary digits.
To overcome any leakage defects of the fluorescent screen, it is necessary to regenerate the charge pattern at a frequency greater than 5 Hz. This is achieved by allowing the amplified output pulses to control the voltage waveforms applied to the modulator and focusing electrodes, while the beam is being scanned. By this means, the information inserted into the storage system can be maintained indefinitely. When the information is required for some other part of a calculation in the computer, the amplified output pulses can be re-routed to the appropriate circuits. It is customary to regenerate and abstract the necessary digits on alternate scans.
The presence of a 'phoney' is indicated by a negative signal in the output lead at some random position on the storage surface. If the amplitude of this signal is too large it may be impossible to obtain a positive signal at this particular position during storage, and the tube is therefore unreliable.
It was originally considered that 'phoneys' were due to two factors: (a) particles of carbon from the internal conductive coating settling on the screen, and (b) pinholes in the screen allowing the electron beam to bombard the glass, which has a different secondary emission coefficient from that of the screen. Using silver instead of graphite as the internal conductive coating eliminated the first of these factors, but the second was shown by experiment to be unconnected with the occurrence of 'phoneys'. The removal of these 'phoneys' has been finally achieved by extremely careful preparation of the phosphor, and by absolute cleanliness while the screen is being deposited and until the bulb with its electrode system is sealed to the pump for evacuation.
In view of the difficulties encountered in preparing good screens, the possibility was considered of using insulated surfaces other than fluorescent powders, but no large-scale investigation into their preparation was made. Although such surfaces, produced by evaporation, would probably have yielded more uniform results, this advantage was outweighed by the desirable property of tubes with fluorescent screens acting as their own monitors and giving a bright display on the screen.
Atmospheric conditions, eg, humidity, have been known to affect the operation of the tubes if they have been made of low-resistivity glass, such as soda glass.
For this reason, high-resistivity lead glass is now used exclusively for the bulbs.
To improve the total storage capacity of the tube it was necessary to reduce the tube spot size. The main methods of achieving this were either to increase the tube EHT or to redesign the electron gun to give an inherently smaller spot without increasing the tube EHT. Either of these methods has the effect of increasing the amplitude of any 'phoneys' present. An increase in EHT, however, would also entail an increase in the voltage required to deflect the beam, so that the redesign of the electron gun was chosen as the best solution. A pentode electron gun is used, the first accelerating anode and second anode (focusing electrode) taking no current, while a small aperture in the final anode allows only a small fraction of the current leaving the cathode to arrive at the screen. The spot size in the redesigned tube operating at 1 kV is about 0.3 mm, and approximately 2,000 digits have been stored on the tube face.
The cathode emission life in these tubes should be no worse than in standard commercial oscilloscope tubes. The rate of deterioration of the storage surface under typical operating conditions has not yet been accurately determined but information on this is being obtained from the behaviour of tubes in the Manchester University computer.