On Sun, 3/13/16, Evan Koblentz via vcf-midatlantic <vcf-midatlantic@lists.vintagecomputerfederation.org> wrote:
I’m starting to conclude the earlier computers were all basically clocked by the media holding the instructions. Is this correct?
I don’t believe so. ENIAC had a master pulse generator.
Brian Stuart (on this list) is making his own ENIAC simulator. Thus he'll have a mich deeper understanding than the rest of us about how the computer really worked.
To follow up on this, one of the 40 racks of the ENIAC is the cycling unit. It's heart is a 100KHz crystal oscillator, but it also has an input and selector switch to be driven from an external oscillator. Every 20 cycles of this main oscillator makes up one addition time and generates a set of 10 clocking signals that are routed around the whole machine to clock all the other units. There's a selector switch to select continuous, one pulse, and one addition modes allowing the operator to single-step either by single cycles of the main oscillator or by single sets of pulses making up one addition time. The cycling unit even has a built-in oscilloscope to display the main clock or any of the 10 signals routed around the machine. IIRC, the cycling unit is one of the units that UPenn has on display in the Moore School building. (Though I may be misremembering and it may be the initiating unit they have.) Generally, if the main storage for the machine was a drum, it made a lot of sense to put the clock on a track or two of the drum and let the drum's motor determine the basic cycle time of the machine. That way you don't have to worry about keeping the internal clock and the drum drive motor synced, and you get your main clock more or less for free. The G-15 we have in the museum works like that. For most other designs, it makes more sense to have a separate clock either based on a crystal or an R/C oscillator. Interestingly, the engineer Harry Huskey was involved with three of the machines mentioned in this discussion, the ENIAC, the Pilot ACE, and the G-15. BLS