100 IN THE BEGINNING

example, the number of flip-flops in the TX-2 is small compared to the gates which transfer information from one group of flip-flops to another. So the flip-flops were allowed to be relatively complicated, but the TX-2 transfer gates were made very simple. A transfer gate is only a single inverter. The emitter is connected to the output of the flip-flop being read, and the collector is connected to the input of the flip- flop being set. The output impedance of the flip-flop is so low that, when the output is at the ground level, a pulse on the base of the transfer gate shorts the input of the other flip-flop to ground and sets its condition.

We planned, of course, to incorporate marginal checking in the design of these circuits so that, under a program of regularly scheduled maintenance, deteriorating components could be located before they caused failure in the system. We also found it practical to use the technique during the design of the circuits to locate the design center of the various parameters and to indicate the tolerance of circuit performance to these parameters. A further application of marginal checking has been found in other systems during shakedown and initial operation to pinpoint noise and other system faults not serious enough to cause failure and therefore very difficult to isolate by other means.

The operating condition of the inverters is indicated by varying the +10 V bias. In the flip flop schematic in Figure 8, the inverters were divided into two groups for marginal checking, and the two leads labeled MCA and MCB were