Accuracy of an Analog Computer

1. To determine how accurate the computing components representing a given variable have to be, assume that they constitute one more piece of measuring equipment which must handle the data before it is fed into an ideal computer. 2. In general, the components of an analog computer must have individual accuracies consistent with the measuring equipment available to the group it belongs to. 3. The static accuracy of an operational amplifier may be measured simply by observing the grid voltage in a closed-loop connection as the output is swung through its full range at some low frequency (Fig. 2). Phase shift may be measured simply (Fig. 3), but usually does not play an important part in establishing limits of performance for the computer. 4. Integrator random drift may be measured simply, but is only one cause contributing to integrator error. Account should be taken of the grid deviations measured in the static test above, or else drift measured about a number of non-zero output levels. 5. The effect of integrator drift is to limit the computing time. Required computing time and highest required frequency are determined directly for real-time simulation. In other computing, one may be traded for the other, so long as the product R remains constant. R is the number of cycles of the highest frequency contained in the longest time. Two computers may be compared by comparing their R's; comparing drift rates, computing times, or maximum frequencies alone is not valid.